In Short
UN38.3 is the United Nations transport-safety test that almost every lithium-ion and lithium-metal cell and battery must pass before it ships by air, sea, rail or road.
- 8 tests (T.1 to T.8): Altitude, thermal cycling, vibration, shock, external short circuit, impact/crush, overcharge and forced discharge.
- Pass criteria: No fire, rupture or disassembly in any sample. T.1 to T.4 also require no leakage or venting and at least 90% voltage retention; T.5 and T.6 add a 170 °C limit.
- Small batteries are not exempt: SP188 simplifies shipping paperwork, not the test. The only narrow exception is SP310 (prototypes and runs of 100 or fewer).
- What it is not: A transport test only. It does not replace UL/IEC product safety standards or an SDS.
Tell a freight forwarder you’re shipping lithium batteries and one of the first questions is “where’s the UN38.3 report?” If it’s missing, outdated, or for a battery that isn’t quite the one in the box, your booking can stall or the cargo can sit at the port. Buyers mostly see the paperwork, the lab sees only the test, and the forwarder sees only the shipping rule, which is how one requirement ends up understood three different ways.
This guide puts those parts together. It covers which batteries need UN38.3, how the sample sequence works, what each of the eight tests checks, why batteries fail, when a design change means retesting, and how UN38.3 differs from an SDS or RoHS.
Table of contents
Which batteries need UN38.3?
Start with the mistake that costs people money: UN38.3 is not a test for big batteries only. A button cell in a blister pack and a large swap-cabinet pack both need it.
It applies to both lithium chemistries, no matter how the battery travels:
- Lithium-ion, the rechargeable kind: LiFePO4 (LFP) and lithium-polymer (LiPo) included, in power tools, e-bikes, motorcycles, swap cabinets, RVs, and energy storage.
- Lithium metal, the non-rechargeable kind: lithium-manganese cells like the CR123A, coin cells, and camera batteries.
- Alone in a box, packed with equipment, or already installed inside it, the test still applies.
The test and the shipping label are two separate decisions. The label depends on the chemistry and how the battery travels:
| Item shipped | UN number |
|---|---|
| Lithium-ion cells/batteries shipped alone | UN3480 |
| Lithium-ion packed with or contained in equipment | UN3481 |
| Lithium-metal cells/batteries shipped alone | UN3090 |
| Lithium-metal packed with or contained in equipment | UN3091 |
In short, almost every commercial lithium cell or battery needs the test. Size only changes how you are allowed to ship it.
A worked example makes the size rule real. A common 36 V 10 Ah e-bike pack is 360 Wh. Its individual cells sit under 20 Wh, but the finished pack is well past the 100 Wh battery line, so that pack ships as regulated Class 9 dangerous goods rather than under the small-battery rules. A loose 18650 replacement cell, around 12 Wh, goes out under the lighter SP188 paperwork. Both still had to pass UN38.3 first.
There is exactly one narrow door for untested stock: special provision 310 (SP310). It covers pre-production prototypes carried for testing, and production runs of 100 cells or batteries or fewer. They still need strong packaging and short-circuit protection. It is a development lane, not a way to send a real product out untested.
For a normal product, the small-battery shipping exception is special provision 188 (SP188). It applies only when every cell and every battery is at or under these limits:
| Chemistry | Per cell | Per battery |
|---|---|---|
| Lithium-ion (incl. LiPo / LiFePO4) | 20 Wh | 100 Wh |
| Lithium metal (primary) | 1 g lithium | 2 g lithium |
Stay under those numbers and SP188 simplifies the box, the marks, and the paperwork. Go over and you ship as full Class 9 dangerous goods, with the stricter packing, state-of-charge, marking, and document rules that come with it. Neither route skips the test; SP188 only lightens the shipping side.
How the UN38.3 test sequence works
A UN38.3 report is not “we tested a battery, here is a certificate.” It is a fixed set of production-representative samples, prepared and counted in a fixed way, and the result applies to that battery design.
Rechargeable products get tested twice. Once on a battery at its very first cycle, fully charged, and again after 25 full charge-discharge cycles. The reason for the second run is simple: a brand-new cell is on its best behavior. The 25-cycle set has to prove the design is still safe after some real use, not just on day one.
How many samples get destroyed depends on size, and the manual draws the size line separately for cells and for batteries (Manual §38.3.2.3, by gross mass). A cell up to 500 g is small, and over 500 g it is large. A battery up to 12 kg is small, and over 12 kg it is large.
For the T.1 to T.5 run on rechargeable products, that means:
- Cells: 5 at first cycle, plus 5 after 25 cycles
- Small batteries: 4 at first cycle, plus 4 after 25 cycles
- Large batteries: 2 at first cycle, plus 2 after 25 cycles
That 500 g / 12 kg line keeps mattering further down. Large cells and batteries soak 12 hours at each T.2 temperature extreme instead of 6. Batteries over 12 kg vibrate at a 2 gn peak rather than 8 gn (gn is a unit of acceleration, where 1 gn is normal gravity). And T.4 shock switches to the mass-based formulas in the table below. Note that these size lines are a different scale from the Wh and gram limits used for SP188 shipping above, so do not use one to guess the other.
Nor are all the samples thrown into one pile. T.1 through T.5 run in order on the same unit, so one battery has to live through all five back to back. T.6 and T.8 use fresh, untouched cells. T.7, the overcharge test, is allowed to reuse a battery that came through T.1 to T.5 unharmed.
If you buy large storage or swap-cabinet packs, this last clause matters. An assembly of 6,200 Wh or less, built from cells that already passed, only needs one assembly sample for T.3, T.4 and T.5 (plus T.7 if it is rechargeable). A larger assembly with proper protection against short circuit, over-discharge, overheating, and overcharge may not need assembly-level testing at all. Nobody has to crush a huge cabinet battery to prove what the cells already proved.
The eight tests (T.1 to T.8): conditions and pass criteria
Here are the eight tests, what actually happens to the battery, and what has to be true afterward. The current rulebook is the 8th Revised Edition (2023), with Amendment 1 adopted in December 2024 and published in 2025. Check the edition printed on any report before you argue over a number, because the limits are tied to it.
| Test | What is done | Pass requirement |
|---|---|---|
| T.1 Altitude simulation | Held at ≤11.6 kPa for at least 6 h at 20 ± 5 °C, simulating an unpressurized cargo hold | No leakage, venting, disassembly, rupture or fire; open-circuit voltage (OCV) ≥90% of pre-test |
| T.2 Thermal test | 10 cycles between +72 ± 2 °C and −40 ± 2 °C, at least 6 h at each extreme for cells and small batteries (12 h for large cells and large batteries), ≤30 min between extremes, then 24 h at ambient | Same as T.1, including OCV ≥90% |
| T.3 Vibration | Sinusoidal logarithmic sweep 7 to 200 Hz and back in 15 min, repeated 12 times per axis on three axes (about 9 h total). Cells and small batteries (≤12 kg) reach an 8 gn peak; large batteries over 12 kg are swept at 2 gn | No leakage, venting, disassembly, rupture or fire; OCV ≥90% |
| T.4 Shock | Half-sine shocks, three in each direction on three axes (18 total). Cells: 150 gn, 6 ms (large cells may instead use 50 gn, 11 ms). Small batteries (≤12 kg): the lesser of 150 gn or √(100850 ÷ mass) gn, 6 ms. Large batteries (>12 kg): the lesser of 50 gn or √(30000 ÷ mass) gn, 11 ms (mass in kg) | Same as T.1 |
| T.5 External short circuit | Case stabilized at 55 ± 2 °C, then shorted through <0.1 Ω, held for at least 1 h after the case returns to temperature (some older reports show 57 ± 4 °C) | Case temperature ≤170 °C; no disassembly, rupture or fire during the test and the following 6 h |
| T.6 Impact / crush | Cells only. Cylindrical cells 18 mm or larger in diameter: a 9.1 kg mass dropped 61 cm onto a 15.8 mm bar. Prismatic, pouch, coin and smaller cylindrical cells: crushed to 13 kN; prismatic cells are tested on two axes | Temperature ≤170 °C; no disassembly or fire during the test and the following 6 h |
| T.7 Overcharge | Batteries and single-cell batteries charged at 2× the maker’s maximum continuous charge current for 24 h; test voltage is the lesser of 2× max charge voltage or 22 V (≤18 V systems), or 1.2× max voltage (>18 V systems) | No disassembly or fire during the test and the following 7 days |
| T.8 Forced discharge | Cells only, driven in reverse from a 12 V supply at the maker’s maximum discharge current, for a time equal to rated capacity ÷ initial current | No disassembly or fire during the test and the following 7 days |
Two rows get misquoted often, and they are the two a careful buyer goes straight to.
T.4 shock is not one flat number. Cells take a fixed 150 gn at 6 ms, but batteries follow their mass, in kg. A small battery takes the lesser of 150 gn and the square root of (100850 divided by mass), and a large battery the lesser of 50 gn and the square root of (30000 divided by mass). The formula starts pulling the figure down above roughly 4.5 kg for small batteries, which is why a 200 kg station pack sees only about 12 gn instead of the 50 gn ceiling. Do not flatten the size rule into “everything under 500 g is small” either: 500 g is the line for cells, and finished batteries use 12 kg.
T.6 is a cell test, full stop. Your finished pack is not getting crushed. You inherit the cell maker’s impact and crush result, which is exactly why you ask for the cell-level report and not just a glossy pack certificate.
UN38.3 is only the transport gate. For the rest of what a battery should survive, IEC 62133-2, IEC 62619, the UL standards, and the pack-level functional and abuse tests, see how they fit together in our lithium battery pack testing guide.
Why batteries fail UN38.3
These are general patterns, not failure statistics. For any real failure, the cause has to be confirmed from the failed sample and its test record. Use the table to see where each test usually points.
| Test | What usually goes wrong | Where the fix is |
|---|---|---|
| T.1, T.2 Altitude and temperature | Voltage drops below the 90% retention limit, or the cell swells or the seal fails after the pressure and temperature swings | Sealing quality, electrode consistency, electrolyte handling |
| T.3, T.4 Vibration and shock | Internal tab or busbar welds separate, cells are not held firmly, fasteners work loose, or interconnects fatigue into a short. These are mechanical and welding faults, not bad chemistry | Rigid fixing, welds with strain relief, real clearance between conductive parts |
| T.5 External short circuit | The protection board is slow to open, a MOSFET or fuse is sized too close to the edge, a solder joint is cold, or creepage distance lets a hot path form. An out-of-balance cell takes more of the stress | Protection electronics and current path, plus cell consistency |
| T.6 Impact and crush | The bare cell fails, because the test is run on the cell, not the pack | Choose a different cell. It cannot be fixed on the assembly line, so ask for the cell-level report |
| T.7 Overcharge | The protection cannot get through 24 hours at 2× the rated charge current without thermal runaway | Cutoff threshold, BMS (battery management system) response, voltage rating of the switching parts, and whether a redundant cutoff exists |
The procurement lesson is simple and worth real money. The cheapest place to pass UN38.3 is at the design and cell-selection stage, before any tooling is cut. Fail a type test and you are paying for a redesign, fresh samples, and a whole new cycle, not just a retest fee.
When retesting is required
A UN38.3 report is not a lifetime passport for your product line—it is a snapshot of one exact build. If you change what’s inside the case, that PDF in your inbox may already be obsolete.
You trigger a retest (or at minimum, a formal engineering reassessment) whenever a modification could alter how the pack behaves under mechanical or electrical abuse. Common real-world tripwires include:
- Cell-level swaps: Changing cell suppliers, swapping chemistries (e.g., NMC to LFP), or switching to a different cell model even within the same brand.
- Spec upgrades: Bumping the Ah capacity, altering nominal pack voltage, or increasing the Wh rating.
- Mechanical & structural tweaks: Redesigning the enclosure, altering how cells are braced/glued, or changing the busbar thickness and spot-weld geometry.
- BMS & protection logic: Changing MOSFET ratings, updating cutoff thresholds, swapping the protection board, or adjusting the overcharge circuit.
- Aftermarket & repairs: Rebuilding packs with non-OEM parts or refurbished cells (an explicit retest trigger reinforced under current regulations).
Does a UN38.3 report expire?
There is no universal expiration date stamped on the certificate. A report doesn't expire just because a year passed; it expires the moment the battery you hand to the freight forwarder stops matching the bill of materials in that test file.
UN38.3 vs MSDS/SDS vs RoHS: which does what
These three documents turn up in the same supplier email and get treated as interchangeable. They are not. Each answers a different question.
| Document | Purpose | Needed to book a shipment? |
|---|---|---|
| UN38.3 test report / Test Summary | Proof the cell or battery passed the eight transport-safety tests; records model, lab, report ID, edition and results | Yes, for applicable lithium batteries, small ones under SP188 included |
| MSDS / SDS | Chemical hazards, handling, storage, emergency and transport-classification information for the product | Usually required as part of the shipping and compliance document set |
| RoHS / REACH | Environmental and market-access rules on restricted substances in products sold into a market | Not a transport or dangerous-goods document, and no substitute for UN38.3 |
The UN38.3 Test Summary is the standard short document a manufacturer or distributor has to make available so a shipper can verify the test. It lists who made the battery, the model, the laboratory, the unique report number and date, the battery description, its watt-hour or lithium content, the test results, the manual edition, and the signatory. You do not need a copy in every box; an electronic version is acceptable. For where the UN numbers and transport data live in the chemical paperwork, our battery MSDS/SDS guide walks through it.
FAQ
Do CR123A batteries need UN38.3?
Yes. A CR123A is a primary lithium-metal (lithium-manganese dioxide) cell, not a lithium-ion cell, so it still requires UN38.3 and ships under UN3090 (alone) or UN3091 (with or in equipment). Confirm the brand’s test summary covers that exact model, since primary cells have their own sample and discharge provisions.
Are HobbyKing / LiPo batteries UN38.3 certified?
A LiPo pack is a rechargeable lithium-ion battery (UN3480/3481) and should have UN38.3 test evidence from the actual manufacturer. “HobbyKing” is a retailer brand, so ask for the test summary and check that the manufacturer and exact model on the document match the cells you receive. A retail listing alone is not proof.
Do small batteries that qualify for special provision 188 still need UN38.3?
Yes. SP188 applies up to 20 Wh per lithium-ion cell and 100 Wh per battery, or 1 g lithium per lithium-metal cell and 2 g per battery. It relaxes the packaging, marking and documentation, but it presupposes that the cell or battery type has already passed UN38.3. There is no size below which the test itself is optional for a normal product.
What about untested prototypes or very small pilot runs?
These fall under special provision 310. A pre-production prototype carried for testing, or a production run of 100 cells or batteries or fewer, may move without a completed UN38.3 report, but it still needs strong packaging and short-circuit protection. It is for development and pilot shipments, not regular sales inventory.
How do I ship batteries under UN38.3?
Passing the test is only the entry condition. You still have to classify the shipment (UN3480/3481/3090/3091), use the correct packaging and lithium mark, observe state-of-charge and quantity limits for the mode, provide the Test Summary and SDS, and follow the airline or ocean-carrier rule. UN38.3 is necessary but not sufficient for a compliant shipment.
Get UN38.3-ready packs for your project
If you are sourcing lithium-ion or LiFePO4 packs, or batteries for swap cabinets, and you need them tested and documented for transport, that is what our custom and OEM lithium battery service is built for. Tell us the model and we will put together the test summary and specification pack. Contact us to start.
Sources
- UNECE, Manual of Tests and Criteria, 8th Revised Edition (2023), Part III, section 38.3 (definitions §38.3.2.3, sample allocation §38.3.3, test procedures §38.3.4). Official free editions index: unece.org/publications/transport/dangerous-goods
- UNECE, Manual of Tests and Criteria, 8th Revised Edition, Amendment 1 (ST/SG/AC.10/11/Rev.8/Amend.1, 2025): ST-SG-AC10-11-Rev8-Amend1e.pdf
- UNECE, Manual of Tests and Criteria, 6th Revised Edition, Amendment 1, §38.3.3. Source for the 25-cycle conditioning and the 5/4/2 sample counts: ST-SG-AC10-11-Rev6-Amend1e.pdf
- UNECE, Model Regulations on the Transport of Dangerous Goods, special provisions 188 (small-battery shipping limits: Li-ion 20 Wh cell / 100 Wh battery; lithium-metal 1 g cell / 2 g battery) and 310 (prototypes for testing and production runs of 100 or fewer): unece.org/transport/dangerous-goods; SP310 interpretive paper: ST-SG-AC.10-C.3-2022-32e.pdf
- IATA, Lithium Battery Guidance Document (current edition): IATA lithium-battery-guidance-document.pdf
