TL;DR: Selecting the wrong standard for your series/parallel configuration isn’t a paperwork problem — it’s a design constraint that forces costly hardware revisions late in development.
TL;DR: Under IEC 62619:2022 clause 8.3, a 4S2P LFP pack must pass a forced internal short circuit test at 100% SOC, a requirement absent from UN38.3 — meaning packs certified only to UN38.3 can legally ship but cannot be deployed in stationary storage in the EU.
What the Standard Scope Boundaries Actually Mean for Multi-Cell Configurations #
Series and parallel configurations are where standards diverge fastest. The common mistake is treating certification as a single pass/fail gate rather than a layered requirement set with application-specific scope boundaries.
Three observable symptoms tell us a design team is operating with the wrong standard framework:
Symptom 1 — Certification obtained, but market access blocked. A pack carries UN38.3 and passes transport without issue, then fails EU market entry for a stationary residential application because IEC 62619 was never initiated. The buyer assumed “certified” meant universally compliant.
Symptom 2 — BMS protection thresholds set to the wrong reference. Cell-level OVP, UVP, and OTP limits are tuned to a GB/T 18287 or GB/T 31241 reference (common when the Shenzhen-area pack house sets firmware defaults) rather than the UL 1973 or IEC 62619 application thresholds. The numbers are close enough that no one notices during development, then a field incident triggers a forensic audit.
Symptom 3 — Test reports show cell-level pass but pack-level gap. The supplier presents a cell certification from EVE or CATL. No pack-level test data exists. These are legally separate certifications, and pack topology — specifically series cell count and parallel string count — changes the abuse test severity and the BMS redundancy expectations.
The diagnostic question to ask upfront is not “which standard do I need?” but “which standards are mandatory for my deployment jurisdiction, and which are mandatory for transport regardless of jurisdiction?”
| Standard | Scope | Applies To | Mandatory Jurisdiction |
|---|---|---|---|
| UN38.3 (Rev. 7) | Transport safety | All Li cells/packs shipped by air, sea, road | Global (IATA, IMDG, ADR) |
| IEC 62619:2022 | Stationary/industrial use | BESS, UPS, telecom backup | EU (EN 62619), China (mandatory since 2023) |
| UL 9540A:2023 | Thermal runaway propagation | BESS installations | USA (AHJ-required in most states) |
| UL 1973:2022 | Stationary/light EV batteries | Racks, cabinets, modular BESS | USA, Canada (ULC S5001 equivalent) |
| GB/T 36276-2023 | Stationary Li-ion BESS | Grid-connected BESS in China | China (mandatory) |
| UN38.3 + IEC 62133-2 | Portable consumer | Portable power stations, consumer packs | EU (EN 62133), required for CE |
The transport/stationary divide is non-negotiable. Every configuration ships under UN38.3. What happens after delivery is determined by deployment class.
The Standard Interaction Layer Most Teams Misdiagnose #
The failure mode here is treating standards as independent silos. They aren’t. IEC 62619:2022 explicitly references IEC 62133-2 for cell-level qualification and pulls UN38.3 transport requirements in by scope. UL 1973 cross-references UL 9540A for installation-level thermal propagation testing. Getting a UL 1973 listing without UL 9540A data is possible at the component level — but the moment you submit to an Authority Having Jurisdiction (AHJ) for an installed system permit in California, New York, or most of the EU, the AHJ will ask for the UL 9540A propagation test data for your specific pack configuration.
This is the non-obvious failure that gets misdiagnosed as a “certification gap” when it’s actually a test sequence problem. Teams certify the cell, certify the pack, obtain transport approval, then discover the installation-level propagation test requires a full rack assembly at final topology. For a 4S4P configuration that becomes 4S8P in the deployed rack, the 9540A test must reflect the deployed configuration, not the shipping unit. Two separate test events, separate cost centers, and the second one can only happen once hardware is finalized.
Measurement method for confirming configuration-specific compliance status: run our internal CVR-09 compliance verification review against the deployed topology as early as the schematic capture stage. The review checks whether the certified unit matches the installed assembly in terms of cell count, string count, and BMS topology. Any mismatch flags a retest requirement. The threshold for a re-test trigger is any change in parallel string count of ≥2 or series cell count of ≥1 from the tested configuration.
Charge-balancing architecture matters here too. A 4S2P pack with passive balancing at 45mA per cell gets through IEC 62619 abuse tests, but if you scale to 8S2P for a 25.6V nominal system, the imbalance accumulation at the end of 300 cycles creates OVP trip conditions that weren’t present in the smaller tested configuration. The standard doesn’t explicitly prohibit this — it just doesn’t test for it. That’s the gap between “compliant on paper” and “compliant under real deployment stress.”
For a detailed breakdown of how BMS balancing thresholds interact with compliance test outcomes, the BMS Engineering documentation covers balancing current minimums, SOC algorithm validation, and protection redundancy requirements across IEC and UL frameworks.
Corrective Actions Ranked by Impact and Feasibility #
-
Establish deployment jurisdiction before any certification is initiated. This sounds obvious; it isn’t practiced. We’ve reviewed RFQs from buyers in 2024 where the standard column was blank. Blank means no specification, which means the factory defaults to whatever their existing test reports cover. Map final deployment country first, then work backwards to mandatory standards.
-
Request the pack-level test report, not just the cell-level. For a Shenzhen-based pack house sourcing EVE 280Ah prismatic cells, a Grade-A cell certificate is standard. Pack-level IEC 62619 or UL 1973 testing is not. Ask specifically for the pack-level report with the tested configuration’s series/parallel topology documented. Absence of a topology-specific pack report means the certification has not been validated for your assembly. This step costs nothing and reveals supplier capability within 48 hours.
-
Align BMS firmware protection thresholds to the standard’s test conditions, not the cell datasheet alone. IEC 62619 clause 7.3 sets OVP test voltage at 1.2× nominal maximum cell voltage. Your BMS OVP threshold must trip before that test voltage is reached. Many Dongguan BMS manufacturers ship firmware with OVP set to the cell datasheet maximum plus 50mV — a delta that can fail the abuse test. Fixing this is a firmware change, typically $0 to $800 in NRE depending on whether the supplier controls their own firmware.
-
Commission a UL 9540A test at the installed rack topology, not the module level. This is the expensive and slow option: $18,000–$35,000 for a full propagation test event depending on test lab and configuration complexity. Schedule it at hardware freeze, not post-production. Moving it earlier reduces risk even if it slightly increases cost.
-
Cross-check GB/T 36276-2023 requirements if any portion of the supply chain or end market touches China. The 2023 revision tightened SOC accuracy requirements to ±5% across the full temperature range (0°C to 45°C) and added mandatory cell-level fusing for parallel strings above 3P. If your pack will be sold into China or manufactured under a Chinese OEM agreement, GB/T 36276 compliance is not optional.
Prevention — What to Specify Upfront #
Put the certification matrix in the purchase order. Not as a reference document — as an acceptance criterion. Line items should specify: UN38.3 Rev. 7 (transport), IEC 62619:2022 or UL 1973 (application, jurisdiction-specific), and IEC 62133-2 at the cell level. State the tested topology by series count and parallel string count. Require that the test report serial numbers match the shipped configuration’s bill of materials, not a generic platform certificate. For Safety & Certification documentation covering certificate verification procedures, incoming inspection steps, and what shared certificates actually reveal about a supplier’s testing history, that reference covers the audit protocol in detail.
The document to request from your supplier before any tooling deposit is the conformity test report with tested topology, test lab accreditation number, and test date within 36 months.
Sourcing Guidance for Buyers #
When evaluating Chinese suppliers for series/parallel configured packs, the first document to request is the IEC 62619 or UL 1973 pack-level test report with the topology explicitly stated. If the supplier provides a cell certificate and calls it a pack certification, that is a qualification gap, not a documentation gap — it means pack-level abuse testing has not been performed for your configuration.
The qualification red flag specific to this category: suppliers who quote a certified “platform” that covers all their series/parallel variants under one report. IEC 62619 and UL 1973 both tie test severity to the specific configuration tested. A 2S4P report does not cover a 4S4P pack. Any supplier who claims otherwise either doesn’t understand the standard scope or is hoping you don’t.
Practical incoming inspection step: pull 3 units from each incoming lot of 50 or more and verify BMS OVP trip voltage under controlled overcharge at 0.2C. Trip should occur at ≤110% of the rated cell maximum voltage. If it trips above 113%, the firmware protection thresholds are misaligned with IEC 62619 clause 7.3 test conditions, regardless of what the certificate says. This takes roughly 4 hours per unit with a benchtop power supply and a calibrated voltage logger.
Published by compactbess.com Technical Team | Request a sourcing consultation
