Same Batch, Seven EL Tests, Defect Rates from 1% to 50%: Where EL Reading Disputes Come From and How to Close Them

ⓘ This article is currently published in Chinese. An English version is being prepared — read the Chinese original.

In a publicly documented solar module dispute, the same batch of modules was EL-tested seven times by broadly similar third-party labs within eighteen months—and the reported defect rate swung from 1% to 50%. That spread is not noise; it is the symptom of an EL test that was never made reproducible. This article takes apart why two honest EL inspections of the same modules can disagree so wildly: different sampling rules, different thresholds for what counts as a crack or a dark cell, different current-injection and exposure settings, the subjectivity of human reading, and the simple fact that cracks grow over time. It then lays out what it takes for an EL result to survive a re-test and hold up as evidence: fixed and documented imaging parameters, a single agreed defect-grading rule, metrological traceability, and a report that lets a third party reproduce the finding rather than just trust it. The point of EL is not to capture one picture—it is to capture a picture someone else can reproduce.