TL;DR: A cell format COA is only as useful as the test conditions behind it — always request the full characterization sheet, not just the summary page.
TL;DR: In our incoming inspection of 31 LFP prismatic lots over 22 months, 11 failed capacity verification at 0.5C discharge against supplier-quoted values, a 35% reject rate that would have gone undetected without lot-level testing.
What a COA Must Actually Contain — And What Most Don’t #
A Certificate of Analysis from a Chinese cell supplier looks authoritative. Serial numbers, discharge curves, QR codes, official stamps. The problem is the fields that are missing, not the ones that are present.
We review incoming COAs under what we internally call the F-07 COA sufficiency gate. This gate has 14 required fields. The average COA from a mid-tier Shenzhen cell supplier passes 9 of them. The five that fail most often: actual measured capacity per cell (not nominal), internal resistance at the exact temperature tested, cycle life test conditions (rate, depth of discharge, cutoff voltage), self-discharge rate measured at 25°C over 28 days, and cell-level UN38.3 test report reference number traceable to the specific cell format and chemistry — not a generic lot.
That last one is where the compliance picture gets ugly. UN38.3 requires testing at the cell configuration level. A cylindrical 21700 test report does not cover a prismatic 280Ah cell, even from the same manufacturer. Yet we’ve received COAs citing a single UN38.3 report number across multiple form factors. When we traced one of those report numbers back through the test lab registry, it covered a 3.2Ah cell. The pack it was attached to was spec’d at 280Ah prismatic LFP. Walk away from that supplier.
Parameters That Predict Cell Grade Before You Run a Single Test #
Four parameters, measured correctly, tell you whether you’re holding a Grade-A cell or a Grade-B reject before you put it on the cycler.
Open-circuit voltage at rest, measured 2 hours post-charge to 3.65V, should read between 3.271V and 3.289V for a properly balanced LFP cell at 25°C. Cells outside that band have either been partially discharged in transit (acceptable if documented) or have elevated self-discharge (not acceptable). The COA should state the measured OCV per cell, not a range. If it gives a range, you’re looking at batch averaging, which means individual outliers are being hidden.
Internal resistance is the second filter. For a 280Ah LFP prismatic at 50% SOC, 25°C, AC impedance at 1 kHz should be below 0.28 mΩ. Grade-B cells or cells with micro-crack electrode damage routinely come in at 0.34–0.41 mΩ. That spread doesn’t sound large, but at pack level across a 16S2P configuration, it drives measurable imbalance and accelerates capacity fade after roughly 400 cycles.
| Parameter | Grade-A Threshold | Grade-B / Reject Range | Test Condition |
|---|---|---|---|
| OCV at rest | 3.271–3.289V | <3.265V or >3.295V | 2h post-charge, 25°C |
| Internal resistance (AC 1kHz) | ≤0.28 mΩ | >0.33 mΩ | 50% SOC, 25°C |
| Capacity retention (0.5C/0.5C) | ≥92% at 500 cycles | <88% at 500 cycles | 2.5–3.65V cutoff |
| Self-discharge rate | ≤1.5% over 28 days | >2.8% over 28 days | 25°C open circuit |
Capacity retention deserves a separate note. The IEC 61960-3 standard sets the test framework for secondary lithium cell characterization, but the test conditions in the COA matter enormously. Suppliers who quote retention figures at 1/3C (0.33C) are giving you the best-case number. Real portable power station loads run at 0.5C to 1C continuous. In our cycler data, LFP cells from three Dongguan-based pack houses showed an average 6.3% lower retention at 500 cycles when tested at 1C versus 0.33C. That gap is buried in most COAs.
The parameter most commonly overlooked by buyers is self-discharge rate. It’s slow to measure (minimum 28 days), which is why many suppliers skip it or substitute a 7-day accelerated figure. Those are not equivalent. A cell that shows 0.9% loss in 7 days can show 3.1% over 28 days due to nonlinear parasitic drain behavior in early SEI stabilization. We require 28-day data per IEEE 1725 as a minimum for any cell entering our approved vendor list.
Decision Framework — Which Format Qualification Path Applies to Your Application #
If you’re sourcing cylindrical 18650 or 21700 cells for a consumer-grade portable power station below 1kWh, the qualification bar is different from what applies to a prismatic 280Ah cell destined for a 100kWh BESS rack. Conflating the two is one of the most expensive mistakes a new procurement team makes.
For cylindrical cells in consumer applications, the incoming inspection sample size should be a minimum of 47 cells per lot (based on AQL 2.5 at inspection level II per IEC 60410 equivalents), with pass/fail based on capacity at 0.5C and internal resistance. If a supplier resists lot-level sampling and insists on COA acceptance only, that resistance is diagnostic. Reputable cylindrical cell suppliers from the Wuxi and Huizhou clusters understand sampling protocols. The ones that push back typically have lot-to-lot variance they’d rather you not see.
For prismatic LFP cells in semi-industrial or stationary applications, qualification shifts to a formation data review requirement. Every Grade-A prismatic cell should come with first-cycle formation data showing coulombic efficiency above 93.7% on cycle 1. That number reflects electrode quality and electrolyte fill consistency. If the supplier can’t provide formation data at all, they either don’t grade by it (a real concern) or they’re sourcing cells from an upstream supplier and re-stamping them (a bigger concern).
If you’re evaluating a supplier for a custom form factor — a pouch cell in a non-standard footprint, for instance — the calculus changes because you’re now assessing tooling maturity, not just current production quality. Tooling age matters: a pouch cell from a die that’s run 2 million+ cycles shows measurably higher tab weld defect rates than tooling under 500,000 cycles, based on our review of 4 custom pouch suppliers in Shenzhen’s Bao’an district in 2023. Request the tooling maintenance log, not just the cell spec sheet. Suppliers who provide it without pushback are in a different reliability tier.
One non-obvious recommendation with a hard boundary: for any prismatic cell order above 500 units, require a third-party capacity verification report from a lab not affiliated with the supplier. This holds for any order where the nominal capacity claim exceeds 100Ah. Below 100Ah, the testing delta between supplier-reported and third-party-measured is typically under 1.8% — within acceptable tolerance. Above 100Ah, we’ve seen discrepancies as large as 7.4%, which translates directly to system design errors downstream. For more context on how cell format affects pack-level design decisions, see our Battery Pack Design section.
Sourcing Guidance for Buyers #
When evaluating Chinese suppliers in this category, the first document to request is the cell formation data summary, not the COA. COAs summarize finished goods. Formation data reveals what happened during manufacturing. A supplier who can produce formation data for the specific lot you’re quoting has traceability infrastructure. A supplier who offers only a generic COA and a brochure does not.
The qualification red flag specific to cell format sourcing is mismatched certification scope. This appears when a supplier’s UL 1642 or IEC 62133 certification covers one cell form factor and they apply it to another for commercial convenience. Cylindrical and prismatic cells are tested as distinct configurations under both standards. If the cert on file covers “cylindrical lithium cells 18650 format” and you’re buying 280Ah prismatic, that cert is decorative.
For incoming inspection, pull a minimum of 12 cells per 500-unit lot and run a 24-hour capacity check at 0.5C discharge from 100% SOC to the manufacturer’s specified cutoff voltage. Reject the lot if more than 2 of 12 cells fall below 97% of nominal capacity. That’s our standard rejection threshold based on pass/fail outcomes correlated against field return rates across 18 months of lot data. For guidance on how cell format selection interacts with BMS protection thresholds, the BMS Engineering section covers this in detail.
Q: Can I accept a COA in place of incoming inspection for a repeat supplier?
A: Repeat supplier status earns a reduced inspection frequency, not inspection elimination. For established suppliers with 4+ consecutive passing lots, we drop sample size to 6 cells per 500 units — but the test still runs. A COA tells you what the supplier measured. Your incoming test tells you what you received.
Q: What’s the most reliable way to distinguish Grade-A from Grade-B LFP prismatic cells?
It depends on which failure mode you’re protecting against. For capacity fraud, first-cycle coulombic efficiency is the clearest discriminator. For long-term reliability, look at internal resistance variance across cells in the same lot. A tight lot (all cells within ±0.04 mΩ of each other) is a stronger Grade-A signal than nominal capacity alone. Grade-B lots often show a bi-modal resistance distribution — two clusters, not one, which means mixed-grade cells were blended.
Q: Do cylindrical and prismatic cells require separate UN38.3 certifications?
Yes. UN38.3 tests are form-factor specific. A prismatic cell must be tested as a prismatic cell. The tests include altitude simulation, thermal, vibration, shock, external short circuit, impact/crush, overcharge, and forced discharge — and the results are geometry-dependent. Any supplier citing a single UN38.3 report across multiple form factors hasn’t run the required tests for at least one of them.
Q: How much should incoming inspection add to per-cell sourcing cost?
For a 500-unit lot with third-party capacity verification on 12 cells, expect $340–$480 for the test run at an independent lab in Shenzhen or Dongguan. On a $0.056/Wh LFP prismatic cell at 280Ah, that’s roughly $0.0009/Wh added to your cost — less than 1.7% of cell cost. Given that a single bad lot at that scale can mean $14,000–$22,000 in rework, the inspection cost is not worth negotiating away.
Published by compactbess.com Technical Team | Request a sourcing consultation
