What a DOT inspector checks on a UN 38.3 test summary for robot battery packs

A UN 38.3 test summary gets a robot battery pack past a carrier's hazmat acceptance review or a DOT inspection when every identifying field on it reconciles with the pack on the pallet and the pack on the shipping paper — model number, manufacturer or applicant, test report ID, the edition of the UN Manual of Tests and Criteria cited, and the list of tests performed. The document does not have to be thick or impressive. It has to survive being compared, field by field, against a label, a packing list, and the shipper's own records by someone who has no reason to give you the benefit of the doubt. Read as a shipping document rather than as a product-compliance souvenir, the summary is the cheapest insurance on the load.

I keep a folder called LITHIUM DOCS that predates most of the people who now ask me for it. Four or five PDFs per pack model, named by part number and revision, plus a photo of the production label. When a new pack model comes through, I build those files before the first pallet goes out, because the alternative is standing at a freight counter while somebody else decides whether your freight moves.

The ninety seconds that matter

The counter is waist high and laminate, with a printer behind it and a clipboard nobody actually uses. The person across it has one job: decide whether the paperwork describes the freight. They are not reading for reassurance. Their first question is not "is this safe." It is "what is this, and does the paper agree with it."

I have been on the wrong side of that exchange with a phone in one hand and a supplier's forwarded PDF open on it, and the review is over in about a minute and a half either way. What buys you those ninety seconds is a summary whose fields you already checked before the pallet was wrapped.

The request itself is almost always the same flat sentence: "Test summary." Under 49 CFR 173.185 and the modal rules that mirror it, a lithium cell or battery offered for transport has to have a UN 38.3 test summary available on request — to carriers and to regulators. Available means available from the offeror, and if you are the one consigning the packs, the offeror is you. Not your contract manufacturer. Not the cell vendor's applications engineer who is somewhere over the Atlantic. You.

So the first thing I fix is not the content of the summary. It is custody of it. I want the PDF, tied to a specific pack revision, with the lab's contact information and the test report ID, sitting in my own document control before the first shipment. Chasing a supplier for a document while a driver stands at your dock is how you find out that "available on request" was a promise about you.

Four fields, then the description

I do the cross-check in the same order every time. Four fields first, because if any of them fails, the rest of the document does not matter.

The model number and description line. Say the summary covers "BP-48-20" and the label on the pallet reads "BP-48-20 Rev C." Those are not the same string. Maybe Rev C is documentation-only and the hardware is identical. Maybe it is not. At the counter, nobody is going to accept your recollection of an engineering change order, and they should not. The same problem shows up as a summary that lists a watt-hour rating or a mass that does not match the label.

The manufacturer or applicant name. This is where cell-level summaries get handed to pack-level buyers and cause real trouble. If the named entity on the summary is the cell maker and the pack was built by someone else, the document is evidence about the cells, not about the pack you are shipping.

The test report ID and its date. The summary is a cover sheet drawn from a full report. That ID is the only thread back to the source document, so it needs to be there, and you need to be able to produce the report if someone asks. It also tells you how old the test campaign actually is.

The test lab's name and address. A summary with no lab identified is a summary with no source.

Only after those four pass do I read the description and the test list in any detail.

The edition line is a date stamp on the rulebook

The summary has to name the edition of the UN Manual of Tests and Criteria the lab tested against, plus any amendments. That line looks like boilerplate and it is not. The Manual runs on revised editions — the sixth came in 2013, the seventh in 2019, the eighth in 2023 — and the test-summary requirement itself arrived in a later revision before being carried into the U.S. hazmat rules through a harmonization cycle.

Two things follow. First, the edition tells you the vintage of the test campaign relative to your design. If the edition cited predates the pack you are shipping, the number on that page is not evidence about your pack. Second, a summary citing an edition so far back that its language does not match what a carrier's acceptance checklist expects is exactly the kind of thing that turns a two-minute review into a call to a supervisor.

Cell summary is not pack summary

The UN 38.3 test summary has to cover the specific battery configuration being shipped — the series and parallel arrangement, the protection circuitry, the way the cells are held and connected — not just the cell inside it. That gap shows up constantly on robot packs, because so many of them are built by a contract manufacturer out of somebody else's cells.

Read the test list with that in mind. T1 through T8 are altitude simulation, thermal, vibration, shock, external short circuit, impact or crush, overcharge, and forced discharge. Not every row applies to every article, and a pack and the cells inside it do not get the same set. Some rows are cell-level by design. A summary that shows one of them as not applicable is not automatically wrong — but I want the applicability stated, not implied, and I want to be able to tell which rows describe the pack and which were inherited from the cell test. If I cannot tell, the document has not done its job, even if the lab did theirs.

What the summary does not do

It does not replace the shipping paper, and it does not replace markings or labels. Those are separate documents with separate failure modes, and passing one has never bought you the other:

• The shipping paper with the correct UN number — UN3480 for batteries shipped alone, UN3481 for batteries packed with or contained in equipment — and the Class 9 lithium battery entry.
• The lithium battery mark with a phone number, where the rules require it on the package.
• The safety data sheet, if the paperwork package for the movement calls for one.
• The state-of-charge limit for air shipments, which lives nowhere on the summary.

I have watched a shipment released on a clean summary and then held twenty minutes later over a missing mark. Same person, two different checks.

The document has no expiration date, which is the actual problem

Nothing on a UN 38.3 test summary ages out. There is no renewal date, no annual review, no clock. A summary issued for a pack built three years ago still reads as valid today, which is why the change question has to be owned by somebody at your company rather than by the document.

What invalidates it: a change of cell vendor or cell part number, a change to the BMS, and any mechanical or electrical change to the pack — a different cell holder, fuse, wiring gauge, connector, potting compound, or a different count of cells in series or parallel. The one that catches people is the quiet one. A supplier sources the same cell specification from a second plant, or accepts an "equivalent" cell to keep a line running, and the summary in your file now describes a pack you no longer receive. Nobody lied. The document simply did not update itself.

I put a document-change clause in the supply agreement for exactly this reason: written notice before any change to cells, the BMS, or pack construction, with the revised summary delivered before the changed part number ships. It is a boring clause. It is also the only thing standing between a re-test cost and a pallet parked at a carrier's dock.

The two-minute test

Before the first shipment of a pack model, and again after any engineering change, I do this: open the summary, the label photo, and the packing list side by side, and try to get from one to the next in two minutes. Model number. Named entity. Report ID. Lab. Edition. Test list. If I cannot, the answer is not to argue with the person at the counter. It is to hold the shipment until the paperwork and the pack agree.

That is the whole discipline. The summary is not a certificate you earn once. It is the first page of a story about one specific pack, and the person across the counter is checking whether the story matches the box.