All chapters
Incoming inspectionIn-process checksEnd-of-line testingTraceability

Quality control on a battery pack is not a final inspection. It is three separate gates — what arrives, what happens during assembly, and what leaves — and the value of each is that it catches something the others cannot. A pack that passes only the last gate has been tested at the point where fixing anything is most expensive.

15.1Incoming cell inspection

Sample per AQL, or 100 per cent for the critical parameters if volumes allow. The purpose is not to duplicate the factory’s work — it is to verify that what arrived is what was ordered, and to build your own record independent of the supplier’s.

Incoming inspection

TestEquipmentAccept criteria
VisualNo dents, leaks, corrosion or damaged terminals. Check the vent area is undisturbed.
Dimensions and weightCalipers, scaleWithin drawing tolerance. Weight catches a wrong or underfilled cell.
OCVMultimeter, 4½ digitBatch spread ≤10 mV. Also confirms arrival SOC is 30–50 %.
ACIR1 kHz milliohm meterBatch spread ≤5 %, and every cell below the datasheet maximum acceptance limit.
Capacity (sample)Cell cyclerAt or above rated, at the datasheet C-rate and temperature.
Self-discharge (sample)OCV logged over 7–14 daysBelow the supplier’s stated K-value limit in mV/day.
Date code and traceabilityRecent, and matching the purchase order.

Important

The self-discharge check is slow and it is the one most often skipped. It is also the only one that catches an internal micro-short before it reaches a customer. If you run one sample test on incoming cells, run this one — see Chapter 4 for why.

15.2In-process checks during assembly

  • Weld quality. Pull-test on samples, and measure resistance across every joint. A busbar joint above roughly 50 µΩ is suspect. At 100 A that is 0.5 W — tolerable; at 1 mΩ it is 10 W, which is not.
  • Sense wire continuity and sequence. The single most common assembly error there is. A sense wire on the wrong tap makes a healthy cell read as dead and trips the pack, and it is far cheaper to catch here than in the field.
  • Insulation resistance. Megger between live parts and the enclosure — typically specified as >500 Ω/V, commonly written as ≥1 MΩ.
  • Hi-pot (dielectric withstand) where the applicable standard requires it.
  • Torque verification with witness marking on every bolted terminal. The witness mark is what lets a later inspection tell whether a joint has moved.
  • Compression on prismatic and pouch modules, within the specified range — a number, not “firm”.

Technical framing

Note the pattern: every one of these is a check on a connection rather than on a cell. That is where assembly-stage defects live, and it matches what actually fails in the field — intermittent faults over bumps, hot joints, and packs that read one cell dead when the cell is fine.

15.3End-of-line pack testing

Ten steps, run on every pack, recorded against its serial number.

  1. 1Full charge to balance. Record the cell delta at top of charge — target ≤30 mV. A pack that will not converge here has a matching or a connection problem, and shipping it guarantees a return.
  2. 2Capacity discharge at the rated C-rate, verified against target. Use the same rate every time so results stay comparable across the product’s life.
  3. 3Pack DCIR measurement — the baseline against which every future field measurement will be judged.
  4. 4BMS protection function test. Actually trigger OVP, UVP, OCP, SCP and OTP and verify each one cuts. A protection that has never been tested is a protection you are assuming.
  5. 5Balancing function verification. Confirm the balancer actually activates, not merely that the board reports it.
  6. 6Communication check — the BMS reports correct cell voltages, current, temperature and SOC over CAN, RS485 or BLE.
  7. 7Thermal check under load — no hot spots, and the gradient across the pack within limit. A thermal camera earns its cost here.
  8. 8IP ingress test on a sample — water spray and dust, to the rating claimed on the label.
  9. 9Vibration on a sample per the AIS or IEC profile that applies.
  10. 10Record everything against the pack serial number.

The bench procedure for the balance-and-capacity portion — equipment, safety, phases, and how to interpret per-cell results — is set out in full in Chapter 2.

15.4The traceability chain

For every pack you ship, you should be able to retrieve:

  • Cell batch and vendor
  • Cell incoming test data
  • BMS serial number and firmware version
  • Assembly operator and date
  • End-of-line test results in full
  • Charger serial number issued with the pack
  • Dealer and customer
  • Every telematics-reported fault since delivery

Important

This is what turns a warranty claim from an argument into a lookup. It is also what lets you spot a bad batch after three failures rather than after a hundred — the same record, read across packs instead of down one.

In plain English

For a business financing the asset, traceability does a third job. When a lender asks why a portfolio is defaulting, “these fourteen packs share a cell batch and a fault signature” is an answer. “We are looking into it” is not, and the difference shows up in your cost of capital.

Quick check: test yourself

1.Which single incoming test would you keep if you could only run one?

Show answer
Self-discharge, measured as OCV drop over 7 to 14 days against the supplier’s K-value limit. It is the slowest and most often skipped, and it is the only incoming test that catches an internal micro-short — the defect that presents months later as one permanently low cell in a customer’s pack. Visual, dimensional, OCV and ACIR checks all catch problems you would likely find at end of line anyway.

2.A busbar joint measures 1 mΩ. The pack draws 100 A continuous. Is that acceptable?

Show answer
No — it is twenty times the sensible limit. P = I²R gives 100² × 0.001 = 10 W dissipated in one joint inside a sealed enclosure. A busbar joint should be under about 50 µΩ, which at the same current is 0.5 W. And it will not stay at 1 mΩ: resistance rises with temperature, so the joint heats, its resistance climbs, and it heats further.

3.Your end-of-line test shows a cell delta of 90 mV at top of charge, against a 30 mV target. Ship it?

Show answer
No. A pack that will not converge at end of line has either a cell matching problem or a connection problem, and both get worse in service rather than settling. Balancing equalises charge but cannot equalise capacity, so a mismatched pack will keep diverging. Diagnose it now — check sense wire seating first, then per-cell capacity — because the alternative is a warranty return with freight in both directions.

4.Why test BMS protections at end of line when the board is certified?

Show answer
Because certification covers the board design, not your assembly of it. A protection can fail to act because a sense wire is on the wrong tap, a temperature probe is dangling in air rather than against a cell, a threshold was left at a factory default meant for a different chemistry, or the firmware shipped with the wrong configuration. Actually triggering OVP, UVP, OCP, SCP and OTP is the only way to know they work in this pack.

Chapter summary

Frequently asked questions

What should incoming cell inspection cover?+

Visual for dents, leaks, corrosion and damaged terminals; dimensions and weight against the drawing; open circuit voltage with a batch spread of 10mV or less, which also confirms arrival state of charge is 30 to 50 per cent; ACIR with a spread of 5 per cent or less and every cell below the datasheet maximum; capacity on a sample at the rated C-rate and temperature; self-discharge on a sample over 7 to 14 days; and date code against the purchase order.

If I could run only one incoming test, which should it be?+

Self-discharge, measured as OCV drop over 7 to 14 days against the supplier’s K-value limit. It is the slowest and most often skipped, and it is the only incoming test that catches an internal micro-short — the defect that presents months later as one permanently low cell in a customer’s pack.

What is an acceptable busbar joint resistance?+

Under about 50 microhms. At 100A that dissipates 0.5W, which is tolerable. At 1 milliohm the same joint dissipates 10W inside a sealed enclosure next to cells, and it will not stay there — resistance rises with temperature, so the joint heats and its resistance climbs further.

Why test BMS protections at end of line if the board is already certified?+

Because certification covers the board design, not your assembly of it. A protection can fail to act because a sense wire is on the wrong tap, a temperature probe is dangling in air rather than against a cell, a threshold was left at a factory default meant for a different chemistry, or the firmware shipped with the wrong configuration. Actually triggering OVP, UVP, OCP, SCP and OTP is the only way to know.

What cell delta should a finished pack show at full charge?+

30mV or less. A pack that will not converge at end of line has either a cell matching problem or a connection problem, and both worsen in service rather than settling — balancing equalises charge but cannot equalise capacity. Diagnose it before shipping, because the alternative is a warranty return with freight in both directions.

Reviewed by

SG

Sahil Goyal

Co-founder, Wingzman

LinkedIn
SG

Sourabh Goyal

Co-founder, Wingzman

LinkedIn

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.