What UN 38.3 Testing Actually Costs, and Why the Lab Fee Is the Small Part
Nearly every lithium battery that ships by air, sea, road or rail has to pass it. The lab fee is the visible line. The sample count, the timeline, and two little-quoted provisions decide what the program actually costs.
The short answer: plan on roughly $2,000 to $10,000 and 4 to 6 weeks for the lab fee on a UN 38.3 program at an accredited Western lab. Published pricing is scarce, offshore labs advertise far less, and the range is consistent with the programs we've seen. On the 28 kWh industrial pack worked through below, the lab fee is about 8% of what the program actually costs; on a 2 kWh pack, closer to a quarter.
The rest is hardware. UN 38.3 requires production-representative packs that are shorted, shaken, overcharged and thermally cycled to destruction: for a large rechargeable battery, eight of them by the Manual's own sample table. And one clause changes the whole equation: a pack over 6,200 Wh assembled from already-tested modules may not need pack-level testing at all. Both halves of that story are below.
- Plan on $2,000–$10,000 and 4–6 weeks in lab fees: roughly 8% of true program cost on a 28 kWh pack, about a quarter on a 2 kWh pack.
- The UN Manual's sample table calls for eight large-battery samples: four through the T.1–T.5 sequence, four for T.7 overcharge. Limited reuse is sometimes permitted; eight is the planning number.
- Most failures happen at T.3 vibration, for mechanical reasons, and a failure costs a fresh sample set, not just a retest fee.
- Under §38.3.3.1(g), a >6,200 Wh pack assembled from tested modules with verified protections may be exempt from pack-level testing.
The line item nobody quotes you
UN 38.3 lives in Part III, sub-section 38.3 of the UN Manual of Tests and Criteria. Section 38.3.3 specifies exactly how many samples you must supply, and the counts depend on whether your battery is “small” (≤12 kg gross) or “large” (>12 kg).
Any meaningful industrial pack is a large battery. The Manual’s sample table, §38.3.3.1 in the current edition, requires, for a rechargeable large battery:
- 2 large batteries at first cycle, fully charged, for tests T.1 through T.5
- 2 large batteries after 25 cycles, fully charged, for the same T.1–T.5 sequence
- 2 large batteries at first cycle plus 2 more after 25 cycles, for the T.7 overcharge test
Eight packs by the table, before spares, before variants, before anything goes wrong. There is one relief valve: the Manual permits undamaged packs from the T.1–T.5 sequence to be reused as T.7’s cycled pair, which can bring a program to six. Ask your lab, and budget as if the answer is no. Tests T.6 and T.8 additionally consume cells, though for most custom pack programs those are covered by the cell manufacturer’s existing cell-level report.
Plan on none of them coming back. T.5 shorts them through less than 0.1 Ω. T.7 charges them at twice the manufacturer’s maximum continuous charge current for 24 hours. What the lab returns is documentation.
One logistics wrinkle worth planning early: those packs are themselves untested lithium batteries, and moving them to the lab is regulated. In the US, the prototype and low-production provision at 49 CFR 173.185(e) is the usual route: ground transport under specific packaging conditions, with air essentially off the table. Build the lab’s location into the plan.
What that costs on a real pack
Take a 28 kWh industrial pack built at $250/kWh, an illustrative, industry-typical figure for a low-to-mid-volume custom build, not a PBC quote. That is $7,000 per pack.
| Line item | Cost |
|---|---|
| Lab fee, UN 38.3 program | $5,000 |
| 8 × 28 kWh packs @ $250/kWh | $56,000 |
| First attempt, all in | $61,000 |
Budget the $5,000 and you have accounted for a twelfth of the program. And this is the optimistic version: it assumes eight packs is enough, that you are certifying one variant, and that you pass on the first attempt.
The same math at 2 kWh
Pack size changes the shares, not the shape. A 2 kWh pack built light enough to stay under the Manual’s 12 kg boundary tests as a small battery: sixteen samples (four at first cycle and four after cycling for the T.1–T.5 sequence, then the same again for T.7), but cheap ones. Build it heavier than 12 kg and it tests as a large battery: eight samples, same as the 28 kWh pack. Using the same illustrative $250/kWh and a lower-end lab fee for the smaller program:
| 2 kWh pack (small, ≤12 kg) | 28 kWh pack (large) | |
|---|---|---|
| Samples required | 16 | 8 |
| Sample hardware | $8,000 (16 × $500) | $56,000 (8 × $7,000) |
| Lab fee (illustrative) | $3,000 | $5,000 |
| Program total | $11,000 | $61,000 |
| Hardware share | ~73% | ~92% |
At the small end the lab fee is a respectable minority of the program; by 28 kWh it is a rounding error. The samples dominate either way, the only question is by how much.
Then there is the failure case
You will most likely fail on vibration. In Compliance Magazine is explicit that most UN 38.3 failures happen during T.3, typically from internal components working loose and breaking interconnects. T.4 shock is second, for the same reason.
That is not a chemistry problem. UN 38.3 is, in practice, a mechanical robustness test wearing an electrical safety costume. Cell retention, interconnect strain relief, weld quality, and bracing of anything heavy are where it is won or lost.
It also compounds, because T.1 through T.5 run in sequence on the same samples. Damage accumulates. A pack that would survive vibration in isolation can still fail the short-circuit test because something shifted three tests earlier.
A failed vibration test does not just cost you a retest fee. A failure normally means a design change, a design change makes a new type, and a new type means a fresh, full sample set.
| Second attempt after a T.3 failure | Cost |
|---|---|
| Retest lab fee | $5,000 |
| 8 replacement packs | $56,000 |
| Design rework, build time, requalification | schedule |
| Running total across two attempts | $122,000 |
This is the argument for spending engineering time on mechanical robustness before samples ship. On a 28 kWh pack, a week of design review that prevents one vibration failure returns about $56,000. There are not many engineering hours with that payback.
The exemption that can zero all of this out
Before you accept the numbers above as your fate, read the least-quoted paragraph in the sub-section. §38.3.3.1(g) of the Manual:
“When batteries that have passed all applicable tests are electrically connected to form a battery in which the aggregate lithium content of all anodes, when fully charged, is more than 500 g, or in the case of a lithium ion battery, with a Watt-hour rating of more than 6200 Wh, the assembled battery does not need to be tested if the assembled battery is of a type that has been verified as preventing: (i) Overcharge; (ii) Short circuits; and (iii) Over discharge between the batteries.”UN Manual of Tests and Criteria, §38.3.3.1(g) in the current edition (§38.3.3(g) in earlier editions)
Read that against the worked example: 28 kWh is 28,000 Wh, more than four times the 6,200 Wh threshold. If the pack is assembled from modules that each carry their own UN 38.3 report, and the assembled design is verified as preventing overcharge, short circuits and over-discharge between those modules, the assembled pack does not need to be tested at all.
The practical read for anyone designing a large system: the certification boundary is an architecture decision. Build a 28 kWh monolith from bare cells and you own eight destroyed packs; build it from tested modules with verified protections between them and the pack-level program may disappear. Module-level testing still has to exist (someone paid for it, often the module vendor), and whether your specific architecture qualifies is a determination to make with your lab, in writing, before you budget. But if you are specifying a large pack and nobody has asked the §38.3.3.1(g) question, ask it. It is worth up to the entire number in the table above.
A companion provision, §38.3.3.1(f), covers assembled batteries at or below 6,200 Wh built from tested batteries: instead of the full program, a reduced test set on a single assembled battery. Between (f) and (g), module-based architectures earn a materially cheaper path through the sub-section at every size.
The eight tests, with actual parameters
The current edition is the 8th Revised Edition (2023), with Amendment 1 issued in 2025. You test against the edition in effect when the design is first tested. These are the real thresholds, not summaries of them.
| Test | What actually happens | Pass criteria |
|---|---|---|
| T.1 Altitude | Held at ≤11.6 kPa for ≥6 hours at 20 ± 5 °C, an unpressurised cargo hold. | No leakage, venting, disassembly, rupture or fire; open-circuit voltage ≥90% of pre-test. |
| T.2 Thermal | 10 cycles between +72 ± 2 °C and −40 ± 2 °C, 6-hour holds (12 hours indicated for large batteries), ≤30 min transitions, then 24 hours ambient. | Same as T.1. |
| T.3 Vibration | Sinusoidal 7–200 Hz, 15-minute sweep repeated 12 times per axis, three axes, 9 hours total. Large batteries to 2 gₙ peak; small to 8 gₙ. | Same as T.1. Most failures occur here. |
| T.4 Shock | 18 half-sine shocks, three positive and three negative per axis. Large batteries: 50 gₙ or √(30000/mass), whichever is smaller, 11 ms; for a 200 kg pack that is about 12 gₙ, not 50. Cells: 150 gₙ, 6 ms. | Same as T.1. |
| T.5 Short circuit | Conditioned to 57 ± 4 °C case temperature, shorted through <0.1 Ω, held ≥1 hour after the case temperature stabilises again. | Case temperature ≤170 °C; no disassembly, rupture or fire during the test or within 6 hours after. |
| T.6 Impact / crush | Cells only. Cylindrical cells ≥18 mm: a 9.1 kg mass dropped 61 cm onto a 15.8 mm bar. Others crushed to 13 kN. | ≤170 °C; no disassembly or fire during the test or within 6 hours. |
| T.7 Overcharge | Charged at 2× the maximum continuous charge current for 24 hours. | No disassembly or fire during the test or within seven days. |
| T.8 Forced discharge | Cells only. Driven in reverse from a 12 V supply at maximum discharge current. | No disassembly or fire during the test or within seven days. |
T.4 is the row summaries get wrong. The 50 gₙ figure is a ceiling, not the test level: the Manual sets large-battery shock at √(30000/mass) gₙ, capped at 50. The formula equals exactly 50 gₙ at the 12 kg boundary and falls as mass rises. Quote a flat “50 g” for a 200 kg pack and you have specified a test four times harsher than the standard asks for.
Three weeks you cannot buy your way out of
Lab queue time is negotiable. The procedure is not. Adding up the Manual’s own dwell and observation requirements for a large battery:
- T.2 thermal: 10 cycles × 12-hour holds at each extreme, plus 24 hours recovery: ten days minimum for a large battery.
- T.3 vibration: 9 hours of shaker time.
- T.5 short circuit: one hour of test plus a six-hour observation window.
- T.7 overcharge: 24 hours plus a seven-day observation period.
That is roughly three weeks of pure test and observation before anyone writes a report. Figure another 10–15 days up front if your samples need the 25-cycle preconditioning. When a lab quotes four to six weeks, most of the gap is scheduling and paperwork, so “can you rush it?” has a much smaller answer than programs assume.
When a design change sends you back
This is the most commercially useful part of the standard and the least discussed. Section 38.3.2.2 defines a “new type” requiring full retesting. For rechargeable batteries: a change in nominal energy of more than 20%, an increase in nominal voltage of more than 20%, or any change that would materially affect the test results.
The operative test is whether a change could lead to failure of any of the tests. The Manual’s accompanying note lists, explicitly non-exhaustively, the kinds of changes that can: changes to electrodes, separator or electrolyte; a change of protective devices, including hardware and software; changes to safety features such as venting valves; a change in the number of component cells; a change in cell connection mode; and mass changes that would affect T.4, since the shock level for large batteries is derived from mass.
The practical read: an enclosure or wiring-harness change is not automatically a new type, but it is not automatically safe either. Mass movement changes the T.4 level; a stiffer enclosure or different cell retention is a T.3 vibration question at identical mass. And the electronics deserve particular respect: because “protective devices, including hardware and software” is named, a BMS change that alters protection behaviour can send you back to the lab. Freeze protection settings before you test. Changing cell count or series-parallel topology sends you back, period. On a 28 kWh pack the new-type boundary is worth $56,000 of hardware, so if you are planning a product family, design the certification boundary, with your lab in writing, before you certify the first variant.
And a question that comes up on nearly every program: a UN 38.3 report does not expire. There is no renewal interval under ADR, IMDG, ICAO/IATA or 49 CFR. What ends its validity is a design change that makes it a new type.
The Test Summary you are required to make available
Since 1 January 2020, manufacturers and distributors must make a UN 38.3 Test Summary available on request. PHMSA harmonised US rules with the requirement at 49 CFR 173.185(a) in its May 2020 rulemaking (HM-215O), with mandatory US compliance from 1 January 2022. The summary has ten mandatory elements, including the testing laboratory’s full contact details, a unique test report identification number, the pack’s mass and watt-hour rating, the test results, the edition of the Manual used, and the name and title of a responsible person; the Manual’s own list adds their signature.
TÜV Rheinland is blunt about what does not count: “Simple informal confirmations in letters or blanket statements in safety data sheets about successful testing… are not sufficient.” One useful concession: IATA’s 2026 guidance confirms you may distribute the summary by QR code or URL rather than on paper with every shipment.
Where it sits in the wider stack
UN 38.3 is transport safety only: it does not make your product saleable. Depending on market and customer you will also need IEC 62133-2 or UL 2054, UL 2271 for light electric vehicles, UL 1973 for stationary and motive auxiliary power, or UL 9540 and UL 9540A for energy storage systems. (How PBC sequences certification inside a pack program is on our technology page.)
The sequencing matters more than the list: UN 38.3 comes first, not because the standards demand an order, but because until it passes you cannot ship product, and often not even samples. IEC and UL programs typically run after or alongside it, and each consumes its own sample set, usually larger than UN 38.3’s: on the order of two to three times the packs, plus cells. If you are certifying a custom battery pack across transport, safety and system standards, the sample hardware alone can run into six figures. That is the number to put in the budget at design freeze, not after.
Budgeting a certification program?
PBC designs packs with the certification boundary, §38.3.3.1(g) included, decided at design freeze, not discovered at the lab.
Frequently asked questions
How much does UN 38.3 testing cost?
Labs rarely publish package pricing; for an accredited Western lab, plan on roughly $2,000 to $10,000, a range consistent with the programs we've seen, and treat it as the smaller line. UN 38.3 destroys the samples it tests: eight for a large rechargeable battery, sixteen for a small one. On a 28 kWh pack built at an illustrative $250/kWh, that is $56,000 of hardware against a $5,000 lab fee: about 92% of the program. On a 2 kWh pack the shares compress: roughly $8,000 of hardware against a lower-end lab fee, so the fee is closer to a quarter of the program.
How many battery packs does UN 38.3 testing consume?
Plan on eight for a large rechargeable battery, meaning any pack over 12 kg gross. The Manual's sample table (§38.3.3.1 in the current edition) requires two large batteries at first cycle and two after 25 cycles for tests T.1 through T.5, plus two more at first cycle and two more after 25 cycles for the T.7 overcharge test. The Manual does permit undamaged packs from the T.1–T.5 sequence to be reused as T.7's cycled pair, which can bring a program to six. Ask your lab, but budget as if it can't. Small batteries (12 kg and under) need sixteen: four at first cycle and four after cycling for T.1–T.5, plus the same again for T.7. Tests T.6 and T.8 additionally consume cells, though these are usually covered by the cell manufacturer's existing report.
How long does UN 38.3 testing take?
Typically 4 to 6 weeks. About three weeks of that is irreducible test and observation time fixed by the procedure: for a large battery the thermal test alone requires ten days, and the overcharge and forced-discharge tests each carry a mandatory seven-day observation period afterward. Figure another 10 to 15 days if your samples need the 25-cycle preconditioning first.
Which UN 38.3 test do batteries fail most often?
T.3 vibration, according to In Compliance Magazine, most commonly from internal components working loose and breaking interconnects. T.4 shock is second, from the same cause. Because tests T.1 through T.5 run in sequence on the same samples, mechanical damage from vibration and shock often surfaces as a failure in a later test. A vibration failure means building and submitting a fresh set of samples, so on a large pack the cost of failing is the hardware, not the retest fee.
Does a battery pack assembled from tested modules need UN 38.3 testing?
Often not. Under section 38.3.3.1(g) of the Manual, a lithium-ion battery with a Watt-hour rating over 6,200 Wh that is assembled from batteries that have already passed all applicable tests does not need to be tested itself, provided the assembled battery's design has been verified as preventing overcharge, short circuits, and over-discharge between the batteries. Whether your architecture qualifies is a determination to make with your lab before you budget for pack-level testing.
Does changing my battery enclosure require UN 38.3 retesting?
Not automatically. But the operative test in section 38.3.2.2 is whether a change could lead to failure of any of the tests, and its example list is explicitly non-exhaustive. An enclosure change that shifts mass affects the T.4 shock level, and one that changes stiffness or cell retention is a T.3 vibration question even at identical mass. Be especially careful with the electronics: a change of protective devices, including hardware and software, is named in the Manual, so a BMS change that alters protection behaviour can make your battery a new type. Cell count, cell connection mode, a change in nominal energy of more than 20%, and an increase in nominal voltage of more than 20% are also named. When in doubt, put the question to your lab in writing.
Does a UN 38.3 test report expire?
No. There is no expiry date or renewal interval for UN 38.3 test reports or test summaries under ADR, IMDG, ICAO/IATA or 49 CFR. A report stops covering your product only when a design change makes it a new type under section 38.3.2.2.
- UN Manual of Tests and Criteria, Part III sub-section 38.3: UNECE Rev.8; PRBA-hosted sub-section text (sample table, exemption and T.4 text quoted from this copy, where they appear at §38.3.3 and §38.3.4.4.2; the current Rev.8 numbering for the sample table is §38.3.3.1)
- TÜV SÜD, Lithium Battery Testing under UN/DOT 38.3
- In Compliance Magazine, Upcoming Changes to UN 38.3 Lithium Battery Testing
- IATA, Lithium Battery Guidance Document, 2026
- TÜV Rheinland, Transport of Lithium Batteries, January 2025
- PHMSA, 49 CFR 173.185, ecfr.gov; test-summary adoption via HM-215O, 85 FR 27810 (May 11, 2020)