UN 38.3 Test Summary Explained: T1 to T8 and the Equipment Behind Each Test

A UN 38.3 test summary is the document that lets a lithium battery travel by air, sea or road. No summary, no shipment. That is the short version, and it is the version most people stop at.

The longer version is where the trouble sits. Summaries we see rejected by freight forwarders are rarely rejected because a test failed. They get rejected because the lab address is missing, the accreditation number is absent, or nobody signed the thing. Meanwhile, behind those eight pass marks sits a rack of equipment that has to hold vacuum, temperature, acceleration and force at the same time, month after month, without drifting.

We build environmental chambers, battery abuse rigs and mechanical test machines in Dongguan. This guide walks through what goes into the summary, what each of the eight tests actually demands from a machine, and what it costs to bring the testing in house instead of waiting weeks for a lab slot.

What a UN 38.3 Test Summary Is, and What It Is Not

Section 38.3 of the UN Manual of Tests and Criteria defines eight tests for lithium cells and batteries. The summary is the two-page document that reports the outcome of those eight tests in a fixed format. It is what a freight forwarder, an airline cargo desk or a port inspector asks for, and it is referenced by the IATA Dangerous Goods Regulations, the IMDG Code and ADR.

It is not a certificate. Nobody "issues" a UN 38.3 certificate, and any supplier who promises one is either simplifying or selling you something that does not exist.

The nine items every summary must contain

This is the part that gets audited, and it is the part that fails. The manual lists the required elements, and a UN 38.3 test summary missing any of them can be refused even when every test passed.

1. Manufacturer name of the cell or battery.

2. Contact details, including address, phone, email and website.

3. Test laboratory name and address, plus its accreditation, normally ISO/IEC 17025 with a certificate number.

4. A unique test report identification number that ties the summary to the underlying report.

5. Date of the test summary.

6. A description of the cell or battery: mass, watt-hour rating, physical description, model numbers and voltages.

7. The list of tests conducted, with a pass or fail result against each of the eight.

8. A reference to the cell test summary when the subject is an assembled battery.

9. A signature, with the printed name and title of the person signing.

Items 3, 4 and 9 cause most of the rejections we hear about. Labs are busy, and a summary produced from a template sometimes ships without the accreditation number or with a scanned signature block that has no name under it.

Summary versus full report

Practical advice: when a new customer asks whether you are "UN 38.3 certified", send both. The summary satisfies their shipping question. The report satisfies the engineer who will otherwise keep asking.

The Eight Tests at a Glance

The tests run in a fixed order, and the samples are not all the same. T1 to T4 may be performed on the same cells. T5 to T8 each need samples that have not been through anything else, because a cell that has already been vibrated or frozen is not a valid starting point for a short-circuit test.

T1 and T2: The Two Environmental Tests

T1 Altitude simulation

Hold the cell at 11.6 kPa or below, which corresponds to roughly 15,000 m, for at least six hours at 20 ± 5 °C. The point is not altitude itself. It is what happens to the cell's internal pressure and to the electrolyte when ambient pressure drops that far.

The machine is a vacuum chamber with three requirements beyond "it pulls vacuum": a pump sized to hold 11.6 kPa with the sample inside, electrical feedthroughs so you can watch voltage and temperature while the door is sealed, and a relief path because a cell that vents at low pressure behaves differently from one that vents at sea level.

What goes wrong in practice is almost always the pump. A chamber that reaches 11.6 kPa empty will drift up when four 5 Ah pouch cells start to outgas. The test then runs at 14 kPa for six hours and nobody notices, because the gauge reading is taken at the start.

T2 Thermal test

Ten cycles. Each cycle holds 72 ± 2 °C for at least six hours, then −40 ± 2 °C for at least six hours, followed by a 24-hour rest at ambient before the next test begins. Ten cycles at that dwell time is not a quick procedure; plan for five to seven days of chamber time per sample set.

Here is the specification detail worth arguing about at the quotation stage. A chamber rated to −40 °C will reach −40 °C empty, on a good day, in a cold room. Put 20 kg of cells inside dissipating heat during charge, and the compressor may only manage −32 °C. Ask for a pull-down curve measured with a heat load, not a brochure number. We have quoted against enough competitors' datasheets to know the gap is real.

T3 and T4: Vibration and Shock

T3 Vibration

The sweep runs from 7 Hz to 200 Hz and back in 15 minutes. That cycle repeats 12 times per axis, on three axes, so three hours of shaking per axis. Below 18 Hz the level is 1 gn. Between 18 and 50 Hz the limit is a 0.8 mm amplitude. From 50 to 200 Hz it is 8 gn.

Buyers spend their attention on the shaker and almost none on the fixture, which is backwards. A fixture with a natural frequency at 180 Hz will amplify the input at that frequency and quietly over-test or under-test the sample. A good fixture is stiff, light, bolted with a defined torque pattern, and has its first mode well above 200 Hz. If a supplier quotes a shaker without discussing the fixture, the test result is a coin flip.

T4 Shock

Eighteen shocks in total: three in each direction along three perpendicular axes, at 150 gn with a 6 ms half-sine pulse. Cells and batteries below 500 g may be tested at a gentler 50 gn and 11 ms, which matters if your product is a coin cell or a small wearable pack.

The machine can be a dedicated shock tester using a dropped table or a pneumatic accelerator, or a shaker with shock capability. What decides whether the result is accepted is the pulse shape. A half-sine that looks like a triangle on the oscilloscope is a failed test setup, not a failed cell, and the tolerance is tight enough that an uncalibrated accelerometer will push you outside it.

T5: External Short Circuit

Short the cell through a circuit with a total resistance below 0.1 Ω, hold it at 57 ± 4 °C, and observe for an hour after the short is applied. The temperature matters more than people expect: a cell that passes at 23 °C can behave very differently at 57 °C, which is roughly the inside of a container sitting on a dock in summer.

The trap is in the word "total". Contact resistance at the terminals, the resistance of the cable, the relay, and the shunt all count. A rig built with 10 mm² cable and a standard industrial relay can sit at 0.15 Ω before the cell is even connected, which invalidates the run. Ask what the measured loop resistance is, and ask how it was measured.

Data logging matters as much as the circuit. You want current, cell case temperature and voltage on the same time base, at a sample rate high enough to catch the first two seconds. That is where the interesting part of the curve lives.

T6: Impact and Crush, the Test Buyers Get Wrong

Impact and crush are two different procedures

The impact test applies to cells. A 15.8 mm diameter bar, weighing 15.1 kg, is dropped from 61 cm onto a cell laid flat on a steel surface. The crush test applies to cells and batteries: the sample is compressed between flat platens to 13 kN, held, and observed.

The force is 13 kN, not 100 kN

This is the most common spec error we see in RFQs. UN 38.3 asks for 13 kN ± 0.78 kN. Chinese EV standards work at a completely different scale: GB 38031 and the older GB/T 31485 discuss forces up to 100 kN, or deformation to 85% of the original dimension.

Both numbers are correct in their own context, and mixing them costs money in one of two directions. Buy a 20 kN machine for UN 38.3 and it will do the job with room to spare. Buy the same machine expecting to run Chinese EV crush tests later and it cannot. If your product roadmap includes traction batteries, specify 100 kN and a platen large enough for a module, then run UN 38.3 at 13 kN on the same frame.

One more detail that separates a usable machine from a legal-looking one: force and displacement must be recorded together. A crush result without a force-displacement curve tells an auditor almost nothing about how the cell actually failed.

T7 and T8: Overcharge and Forced Discharge

T7 Overcharge

Charge at twice the manufacturer's maximum continuous charge current, for 24 hours, while monitoring temperature and voltage. Two things sink cheap equipment here. First, the supply has to hold a constant current into a cell whose voltage is climbing, which means a proper CC-CV controller rather than a bench power supply. Second, 24 hours unattended is a safety question, not just a data question, so the rig needs a temperature cutoff and an independent over-voltage limit that does not depend on the same software doing the charging.

T8 Forced discharge

Force current through the cell at the 1 It A rate, in the reverse direction, typically with cells connected in series against a DC source. A power supply alone will not do it. You need a unit that can sink current as well as source it, which in practice means a bidirectional supply or an electronic load paired with a charger. Ask specifically whether the model can absorb, and at what power. It is a common gap in datasheets.

Pre-Testing In House: What to Buy, in What Order

Third-party labs are competent and, in most of Asia, reasonably priced. The problem is the queue. A full eight-test round runs two to four weeks of lab time, and a failure at T6 or T7 sends you back to the front of that queue with a design change. One failed round costs more in calendar time than the equipment would have cost to buy.

In-house pre-testing does not replace the accredited lab. What it does is let you arrive at the lab with a design that has already survived the eight tests, so the UN 38.3 test summary you pay for gets issued on the first attempt. That is the entire business case, and it holds up in most factories we talk to.

Total for a working line lands somewhere between 50,000 and 130,000 USD depending on force capacity, chamber volume and how much automation you want. Compare that with a single failed certification round: the lab fee, the freight, the re-test, and three to six weeks of a product launch sitting still. Most companies that buy the line do it after their second failure, which is about one failure too late.

UN 38.3 Test Summary FAQ

Is a UN 38.3 test summary the same as a test report?

No. The summary is a short document with nine required elements that accompanies shipments and satisfies carriers. The full report is the detailed record behind it, with curves, photographs and sample identification. Customers and auditors usually want the report; forwarders want the summary. Send both and stop the email chain early.

Does a UN 38.3 test summary expire?

The manual does not set an expiry date. In practice the summary becomes invalid when the cell or battery design, chemistry or manufacturer changes, and buyers usually ask for one issued within the last three years. Some carriers and some customers apply their own age limit, so keep the latest revision on file and reissue after any change to the bill of materials.

Which tests can share the same samples?

T1 to T4 may be run on the same cells, in that order. T5 to T8 each require samples that have not been through any other test. That rule is why a full round consumes more sample cells than people budget for, typically 30 to 60 pieces depending on the product.

Can I run UN 38.3 testing in my own factory and write the summary myself?

You can run the tests, and many manufacturers do. Whether the resulting summary is accepted depends on the lab accreditation behind it. Airlines, forwarders and most large customers expect an ISO/IEC 17025 accredited lab, so in-house testing is normally used for pre-qualification rather than for the document that travels with the goods.

Do battery packs need all eight tests too?

Yes. All eight apply to batteries as well as cells, and the battery summary must reference the cell test summary. If your pack uses cells from a supplier who will not share their summary, that becomes a documentation problem at exactly the wrong moment.

What does a full eight-test round cost at a third-party lab?

In China, roughly 3,000 to 8,000 USD per model for a standard chemistry, with extra sample preparation and expedited scheduling on top. In Europe or North America the same scope usually runs 8,000 to 20,000 USD. Prototype and low-volume exceptions cost more because they involve additional documentation.

What is the most common reason a UN 38.3 test fails?

In our experience, T6 crush and T7 overcharge account for most failures, and vibration failures are usually a fixture problem rather than a cell problem. Before blaming the design, check whether the fixture was stiff enough and whether the pulse shape matched the specified half-sine.