Eight UN 38.3 tests, and whether the summary covers your pack

A UN 38.3 test summary covers your pack only when the configuration it identifies, meaning the cell, BMS or protection circuitry, enclosure, and internal construction, matches the pack you are ordering. If any one of those four items differs from the summary, the eight numbered results describe a different battery, and the summary is not evidence for the build on your purchase order.

Come stand at the table with me. I have the supplier's summary printed on one side, the purchase order on the other, and a pack sample with its lid off on a rubber mat. The lid screws rattle in a paper cup. The summary's configuration table is the only part I read first; the eight rows can wait. I run my finger down the cell line, the BMS line, the enclosure line, and the internal construction note, and I compare each one to the PO and to what is physically in front of me. The cell wrap in the sample carries a different part number than the one in the summary. The enclosure drawing shows an aluminum extrusion; the sample is a molded plastic case with a gasket seam. The BMS is described as "protection circuit" with no board revision. The eight tests may all say pass. None of that matters yet.

The eight rows are results for one build

The summary reports eight numbered tests: altitude, thermal cycling, vibration, shock, external short circuit, impact, overcharge, and forced discharge. I want those rows present. A missing row is an easy stop. But the rows are results for a tested item. The configuration table tells me what that item was.

This is the part that gets lost when a buyer treats the summary like a product-family datasheet. It is not. It is a document about one build. A test summary is issued by the testing party and furnished by the manufacturer, and it should identify the cell, the pack, and the testing laboratory. It is not a certificate. It is a record that points to a tested configuration, and the first question is whether that configuration is the one on your PO.

Capacity and voltage are the easy numbers to compare. They are also the least useful here. Two packs can both be 48 V and 20 Ah and still differ in cell, BMS, enclosure, and internal construction. The summary is about the second set.

The four lines I check before the test rows

I make a short comparison sheet. It has four rows, and each row has a summary field, a PO field, and a physical check. If a row does not line up, I stop.

Cell

I want the cell manufacturer and part number, not just the chemistry. "Lithium-ion, NMC" is not a cell identity. If the summary says one cell part number and the PO says another, the eight rows belong to a different build. I also check the cell format and capacity in the summary against the packing list or sample. A cell change is not a small substitution when the document is supposed to describe the tested article.

BMS or protection circuitry

I look for a board part number and revision, not a generic phrase like "protection circuit." The BMS controls overcharge, overdischarge, short-circuit response, and in many robot packs, communication with the host. If the summary tested one board and the PO ships another, the overcharge and external short-circuit results are not about your pack. I also check whether the summary identifies the protection circuitry as part of the tested pack or only as a cell-level component.

Enclosure

The enclosure is not cosmetic. It affects creepage, spacing, mounting, ingress protection, and how the pack behaves in shock and impact. I compare the summary's enclosure description or drawing revision to the PO drawing and to the sample. If the summary says a metal extrusion and the sample is a molded plastic case, that is not a labeling error I can wave through. It is a different mechanical build.

Internal construction

This is the line buyers skip most often. Internal construction covers cell arrangement, cell holders, busbars, welds, fasteners, potting, insulation, and the spacing between live parts. A summary may not give you a full drawing, but it should identify the construction or reference a build drawing. If the PO changes the cell arrangement from 4S2P to 4S1P, or changes the busbar material, the old summary may no longer describe the pack being shipped.

Cell-level and pack-level are different questions

A cell-level summary and a pack-level summary are not interchangeable. Cell testing answers whether a cell design passed the eight tests as a cell. Pack testing answers whether a pack build passed them as a pack. The pack adds protection circuitry, enclosure, connections, and internal construction. A cell that passed at cell level does not tell me the pack's BMS and enclosure were tested as a pack.

Sometimes a program needs both. Sometimes the supplier provides a cell summary plus a pack summary. What I do not accept is a cell-level summary offered as if it covers the pack. I also do not accept a pack-level summary for a different pack build just because the voltage and capacity match.

When the build changes, the summary ages out

A summary is tied to a moment in the build. When the buyer changes the cell, the BMS, or the mechanical construction, an existing summary may no longer describe the configuration being shipped. The change does not have to be dramatic. A new cell supplier, a BMS revision, a different enclosure molding, a revised busbar, or a new cell holder can all move the pack away from the tested article. I check the issue date against the last configuration change I know about.

The pattern is common enough: capacity and voltage stay the same through a change, and the old summary still looks right on the cover. That is the failure mode I am trying to catch before the PO, not after the first shipment. The summary is not wrong. It is just about the previous build.

What I ask for in writing

I attach the PO configuration to the request and ask for line-by-line confirmation. The language is plain:

"Please confirm that the UN 38.3 test summary provided covers the exact pack configuration in the attached purchase order, including cell part number, BMS board part number and revision, enclosure drawing revision, and internal construction drawing revision. If any of these differ, provide the test summary for the configuration being shipped."

That last sentence matters. If the supplier can only produce a summary for a different pack build, I do not have evidence that the configuration being purchased was tested. I have evidence that something else was. That is not a document review pass.

If the supplier sends a family summary, I ask for the build sheet that ties my configuration to the tested article. A family summary can be workable if the build sheet and the summary reference each other and the four lines match. A verbal 'same design' is not enough. I want the identifier on the page.

What I do with a mismatch

I do not argue about the eight rows first. I put the summary beside the PO, mark the mismatched line, and ask for the summary that matches the build. If the program cannot produce one, the buyer has a choice: change the build to the tested configuration, fund a new test program and summary, or accept that the evidence does not cover the pack. That is a commercial and schedule decision, not a document trick.

The eight tests are important. I still want all eight rows, and I want the testing lab identified. But the first question is narrower: does this summary describe the cell, BMS, enclosure, and internal construction of the pack I am buying? If the answer is no, the eight rows are not about my pack.