TL;DR: The sample request process for cell format evaluation is where most design-in timelines collapse — not because of slow shipping, but because engineers ask the wrong questions upfront and receive datasheets that can’t be validated.
TL;DR: In our evaluation pipeline, cells that pass datasheet review but fail incoming impedance testing at 1kHz account for roughly 34% of rejected sample lots over the past 18 months.
What to Specify Before You Send a Single Inquiry #
Most engineers send a vague inquiry — “we need 21700 cells, 4000mAh, LFP or NMC” — and wonder why they get six unusable quotes back. The inquiry itself is a qualification instrument. What you specify tells the supplier what kind of buyer you are, and it determines whether the technical team or the sales rep responds.
For cylindrical cells (18650, 21700, 26700), your inquiry must specify nominal capacity, minimum discharge capacity at your target C-rate (not 0.2C), and the temperature range for rated performance. A cell rated at 5,000mAh at 0.2C/25°C may deliver 4,340mAh at 1C/25°C and only 3,870mAh at 1C/10°C. If your application runs in an unheated enclosure, that last number is the one that matters. Specify it.
For prismatic LFP cells, also request dimensional tolerances on terminal position and can height, not just overall envelope dimensions. A ±0.8mm terminal offset across a 16S pack introduces measurable busbar stress and has caused cracked weld joints in production builds we’ve reviewed. This isn’t in any datasheet — you have to ask.
For pouch cells, add electrolyte swelling specification (typically maximum 0.2mm increase at 100% SOC after 100 cycles) and tab pull strength (minimum 60N per IEC 62133-2 clause 7.3.2 mechanical integrity test). Suppliers who can answer this without a two-day delay have in-house characterization capability. Suppliers who forward it to their factory contact probably don’t.
The internal document we use for this is our SRQ-04 Sample Request Qualification form, which forces engineers to complete 11 fields before any inquiry goes out. The refusal rate from qualified Shenzhen-based cell pack houses drops significantly when the inquiry arrives pre-filled with actual test conditions.
| Parameter | Cylindrical (21700) | Prismatic LFP (280Ah) | Pouch (NMC, 50Ah) |
|---|---|---|---|
| Capacity test rate to specify | 0.5C and 1C | 0.33C and 0.5C | 0.5C and 1C |
| Temperature conditions | 25°C and 10°C | 25°C and 45°C | 25°C and 0°C |
| Cycle life threshold to request | ≥500 cycles to 80% (1C/1C) | ≥3,500 cycles to 80% (0.5C/0.5C) | ≥800 cycles to 80% (0.5C/0.5C) |
| Critical dimensional spec | Can diameter ±0.05mm | Terminal offset ±0.5mm | Swelling ≤0.2mm at 100 cycles |
| Minimum tab/terminal spec | N/A | Torque spec on terminals | Tab pull ≥60N |
The cycle life thresholds above are what we require suppliers to substantiate with actual test data, not marketing claims. If a prismatic LFP supplier can’t show 3,500-cycle data at 0.5C/0.5C, they’re quoting cycle life from interpolation or from a different cell grade.
What Fails at Incoming Inspection — and the Mechanism Behind Each Failure #
AC impedance mismatch at 1kHz
This is the failure we see most consistently. A supplier ships samples with 4,200mAh capacity confirmed at their facility. Your incoming check agrees on capacity. But when you run 1kHz AC impedance on a full 25-sample incoming lot using IEC 62660-1 test conditions, you find a standard deviation of 4.7mΩ across the lot, with two outliers at 31mΩ versus a lot average of 18mΩ. Those two cells will age at a different rate in any series-connected pack. The BMS can’t compensate for internal resistance spread — it can only respond to voltage, which diverges later. By the time you see cell voltage spread in the field, the outlier cells have already accumulated 200+ additional cycles of stress.
The mechanism: impedance spread at delivery reflects grading inconsistency at the factory, often because the supplier is mixing cells from multiple production runs to fulfill smaller orders. This is common at Dongguan-area pack houses that don’t run their own formation lines and source cells from secondary distributors. The check is straightforward — 1kHz ACIR on 100% of incoming samples for the first three lots, then reduce to AQL 2.5 sampling after the supplier demonstrates lot-to-lot consistency within ±1.5mΩ.
Datasheet capacity at mismatched conditions
A North American portable power station brand sourced 21700 NMC cells in 2023 from a Shenzhen supplier whose datasheet showed 5,000mAh at 0.2C/25°C. Incoming testing at 0.5C/25°C returned 4,614mAh — acceptable. But testing at 0.5C/10°C, which reflected actual winter warehouse conditions during a product recall investigation, returned 3,983mAh. That’s a 20.3% capacity shortfall at a temperature the datasheet never mentioned, and it explained why field units showed “low battery” alerts within 45 minutes of use in cold environments. The batch was 12,000 cells deep before anyone tested at 10°C. The capacity specification in the original inquiry had only referenced 25°C.
The mechanism is straightforward: LFP and NMC cells both show strong temperature-dependent capacity drop, but the curve shape differs. NMC loses capacity more steeply below 10°C than LFP. If your inquiry doesn’t specify the test temperature, the supplier will always quote the favorable number.
Cycle life from accelerated or non-representative protocols
We’ve received cycle life data from at least six suppliers in the past two years where the test rate was 0.2C charge / 0.2C discharge. That produces numbers like 2,847 cycles to 80% for a cell that delivers 1,100 cycles under a real 1C/1C protocol. The UN38.3 Section 38.3.4 cycling test uses specific conditions, but it’s a safety test, not a performance benchmark. Suppliers know this. Requesting cycle life data without specifying charge and discharge C-rate, temperature, and depth of discharge is an open invitation to cherry-pick.
For a battery pack design application where projected service life is three to five years, the minimum meaningful cycle test is 1C/1C at 25°C to 80% capacity retention. Ask for the raw capacity-versus-cycle curve, not just the summary number. A supplier with genuine data will provide it. One without it will send a graph with suspiciously smooth progression and no error bars.
How Many Samples Do You Actually Need? #
For initial electrical characterization, 25 cells is the practical minimum for cylindrical formats. For pouch cells above 20Ah, we typically request 15 units given the cost delta. For prismatic LFP 280Ah cells, 10 units per sample lot is sufficient for capacity, impedance, and dimensional validation — you’re not doing statistical process control at this stage, you’re confirming the supplier can hit spec.
That said, if you’re planning a cycle life test that runs to 500 cycles minimum, request a separate set of 5 cells dedicated to destructive cycling. Don’t pull from your characterization lot. A cell that has been repeatedly charged and discharged for capacity verification has already accumulated partial aging, and mixing it into a cycle life test contaminates your data. Request two distinct sample sets in your inquiry: one for incoming characterization, one for life testing.
Some Shenzhen factories will push back on sending 30+ cells as free samples. That’s fair. Budget $0.80–$1.20 per cylindrical 21700 cell for sample cost reimbursement, or negotiate it against a committed qualification timeline. The cost of getting adequate samples is trivial against the engineering time you’ll waste repeating a test because your initial lot was too small.
Sourcing Guidance for Buyers #
When evaluating Chinese cell suppliers for portable energy storage applications, the first document to request is the Formation & Grading Record for the specific lot you’re sampling — not the generic product datasheet. Its absence signals one of two things: the supplier is a trader rather than a manufacturer, or they’re grading cells from mixed production batches and don’t want you to see the distribution. Either way, it changes how you price the risk.
The qualification red flag specific to cell format evaluation: any supplier who quotes cycle life without specifying test C-rate, temperature, and DOD in the same sentence is giving you a marketing number, not a characterization result. This applies equally to cylindrical, prismatic, and pouch formats.
For incoming inspection on your first two qualification lots, run 100% ACIR at 1kHz and flag any cell where impedance deviates more than 15% from the lot mean. For capacity, sample at minimum 20% of the lot (or 25 cells, whichever is larger) at your actual application discharge rate and temperature. Do not accept supplier test data as a substitute for this step on the first two lots. After that, UL 1973 Annex requirements and your own statistical baseline together give you a defensible AQL threshold for ongoing receipt inspection.
Also document the BMS engineering requirements for the pack design before you finalize cell selection — cell format and BMS architecture interact in ways that aren’t obvious until you’re mid-layout.
Frequently Asked Questions #
How long should the full evaluation-to-design-in process take for a new cell supplier?
Budget 14 to 18 weeks if you’re running a proper qualification. Sample transit from Shenzhen takes 1–2 weeks. Incoming characterization and impedance sorting: 1 week. Accelerated cycle testing to 500 cycles at 1C/1C: roughly 6–7 weeks at two cycles per day with rest periods. Datasheet validation, dimensional audit, and supplier response review add another 2–3 weeks. If someone promises you a fully qualified cell in six weeks, they’re skipping the cycle testing.
Is it acceptable to use supplier-provided test data instead of running your own incoming tests?
It depends on where you are in the qualification process and what your application risk tolerance is. For consumer electronics OEMs where battery failure means a product return, third-party-verified data from a CNAS-accredited lab is the minimum acceptable for high-volume design-in. For early prototype evaluation, supplier data is fine as a screening filter — but you still need your own impedance check on the actual cells you received, because lot-to-lot variation means the datasheet and the physical sample may not represent the same population.
What’s the right way to progress from sample approval to production supply agreement?
Lock three things before signing: the cell grade designation (Grade A, Grade A-, or equivalent in the supplier’s internal classification), the lot-to-lot ACIR tolerance (we specify ±1.5mΩ max standard deviation), and a clear statement on which production line and formation equipment will be used. Suppliers occasionally switch formation equipment between sample production and volume runs. That change, undisclosed, is the primary reason qualified samples outperform production cells — and it’s contractually preventable if you specify it upfront.
Do different cell formats require different test protocols?
Yes, and the differences are non-trivial. Pouch cells require a swelling measurement fixture and a tab mechanical integrity check that simply doesn’t apply to cylindrical or prismatic formats. Prismatic cells require terminal torque verification and dimensional audit of the busbar interface. Cylindrical cells are the most standardized to test but the most prone to impedance spread in large lots. Run the format-appropriate protocol, not a generic one.
Can a small buyer request evaluation samples from Tier-1 Chinese cell manufacturers?
Realistically, CATL and BYD won’t engage directly with buyers under roughly $2M annual projected volume. EVE Energy and CALB have lower thresholds but still require a credible application brief. Below those tiers, there are legitimate Tier-2 and Tier-3 manufacturers in the Shenzhen, Dongguan, and Huizhou clusters producing cells that meet IEC 62619 safety requirements with independently auditable test reports. The access question and the quality question are separate — don’t conflate them.
Published by compactbess.com Technical Team | Request a sourcing consultation
