How to Assess Solar Module I-V Curves and Defect Detection: the Measurement Standards (IEC 60904-3 Reference Spectrum; IEC 60891 Temperature and Irradiance Corrections, Procedure 4 for c-Si With Unknown Coefficients; IEC 61215-2 Qualification Sequence With MQT 20–22 Added in 2021), the Imaging Standards (IEC TS 60904-13 Electroluminescence Under Forward Bias; IEC TS 62446-3 Outdoor Infrared With Its Thermal-Abnormality Matrix; IEC 62446-1 Commissioning Tests), What the Field Evidence Says (Good IR Images Recommended Above 600 W/m² in Stable Conditions; IR First to Localise, Then EL in the Lab or in Near-Dark On Site; Standard Field I-V Tracers Accurate to About 5% and Blind to Hot Spots Not Yet Costing Power) and One Arrival-Inspection Dataset (78% and 70% Major-Crack Rates in Two 50-Module Samples of a Multi-Wire Half-Cell PERC Type, Where a Dendritic Crack Costs ~0.2% of Module Power per Half-Cell)

Short answer. Two standards sit behind any I-V curve offered as evidence: IEC 60904-3, which relates the performance rating to a common reference spectrum, and IEC 60891, which defines the temperature and irradiance corrections applied to a measured curve — including, since 2021, a procedure for crystalline silicon with unknown temperature coefficients. Defect detection then uses two imaging methods with their own technical specifications: electroluminescence under forward bias (IEC TS 60904-13) and outdoor infrared thermography (IEC TS 62446-3, with its matrix of thermal abnormalities), inside the commissioning and inspection frame of IEC 62446-1. The field literature says how each is used: good IR images are recommended above 600 W/m² in stable weather, IR is the suggested first step to localise anomalies, and EL the second — in a lab after unmounting, or on site in near-dark conditions; a standard field I-V tracer is accurate to about 5% and will not see a hot spot that is not yet costing power. One arrival inspection shows what EL finds in practice: 78% and 70% of two 50-module samples had dendritic cracks.

1. The measurement standards

CNWattHub reading: an I-V curve measured away from the reference conditions needs its temperature and irradiance correction procedure stated before it is compared with a rating given at those conditions. The reviewer's first question is which IEC 60891 procedure was applied, and — for a crystalline-silicon module whose temperature coefficients were not supplied — whether procedure 4 was used. The second is which reference spectrum the measurement was related to; IEC 60904-3 supplies that reference and, through it, the spectral basis on which IEC 60904-9 classifies simulators. The IEC 61215-2 qualification data are the design-qualification reference, obtained under the sequence's own cost-and-time constraints. This is a reading of the three abstracts' division of labour; none of the clause texts was read.

2. The imaging standards

CNWattHub reading: the two technical specifications are defined for different conditions. TS 60904-13 applies to modules measured with a power supply forward-biasing the cells — the module is driven rather than illuminated; TS 62446-3 is defined for modules and plants in operation. The field literature's suggested order, IR to localise and then EL (section 3), fits that division. The 'matrix of thermal abnormalities' in TS 62446-3 is the document to ask for when an IR report classifies findings: the classification should map to it.

3. What the field evidence says

The IEA PVPS Task 13 review of IR and EL imaging (2018) is the most direct source on method limits:

• IR conditions. IR detects power-relevant malfunctioning modules, strings and devices if artefacts — partial shading, dirt — can be excluded; for high-quality images the review recommends irradiance above 600 W/m² and constant conditions: no clouds, low wind, stable ambient temperature.
• IR purpose and difficulty. The aim is to catch failures before yield loss or serious damage; because the damage has often not developed yet, objective interpretation 'is not trivial'; hot spots with strong local gradients can significantly affect performance and in extreme cases lead to smouldering fires.
• Order of methods. After unexpectedly low output, IR should be used first to localise anomalous modules, which are then unmounted and sent to a lab for EL — or EL is done on site, without unmounting, in near-dark conditions with specific instrumentation.
• The string effect. A faulty module whose I-V curve differs from its neighbours operates at a different point inside the string from the point it would reach alone under a resistive load; the string condition can be reproduced on a single module with a controlled power source whose MPP current set point is computed from datasheet values at the measured irradiance and temperature.
• Linking IR to the curve. Combining several IR methods lets typical c-Si defects and their influence on the I-V curve (series resistance, short-circuit current) be explained; under sun-illuminated thermography several defects can be told apart by varying the electrical load.

NREL's O&M guide adds the field I-V tracer's limits: manual tests (open-circuit voltage, operating current, I-V tracing) find DC faults the monitoring system cannot; accuracy is bounded by the combined irradiance, temperature and electrical sensors and is about 5% for standard field units; the test shows only defects already causing significant failure and will not detect hot spots that are not yet degrading output electrically. Its template maintenance schedule puts annual I-V traces on a percentage of strings and IR scans on combiner boxes, and it illustrates high-resolution aerial IR identifying failed strings, modules and cells.

CNWattHub reading: the two sources agree from opposite directions. The tracer sees electrical loss but not latent hot spots; IR sees latent hot spots but is 'not trivial' to interpret before loss appears. A defect-detection programme that uses only one of them has a blind spot the other covers; and a single-module I-V curve taken under a resistive load does not show the operating point the module has inside its string, which is why the review describes reproducing the string condition with a controlled source. The 5% tracer figure also sets a floor: a claimed power shortfall smaller than the tracer's own uncertainty is not a finding unless the measurement uncertainty of the specific instrument and conditions has been accounted for.

4. One arrival-inspection dataset

The IEA PVPS Task 13 degradation and failure report (2025) gives a concrete example of what EL finds on arrival, for a multi-wire interconnected half-cell PERC glass/backsheet module type: the same batch was delivered to two sites and 50 modules were EL-inspected at each; defects were classed major (dendritic cracks), minor (simple branch cracks) or normal (a few single cracks). One site: 78% major, 18% minor, 4% normal; the other: 70% major, 30% minor, none normal. The report adds that in this module type a dendritic crack costs only about 0.2% of module power per affected cell, while cracks on many cells can add up to a noticeable loss.

CNWattHub arithmetic (on the tagged figures): of 100 modules inspected across the two sites, 74 were in the major class and 2 in the normal class; yet the stated per-cell power cost of a dendritic crack in this type is ~0.2%, so a crack census and a power census tell different stories about the same batch. That is why the classification matters: an acceptance criterion written in crack classes and one written in measured power can diverge for the same modules, and how far depends on how many cells are cracked and on the thresholds chosen. Which criterion a contract uses is a contract question; this guide only shows that the two can diverge.

5. An assessment sequence built from the sources

1. Measurement basis. Spectral reference (IEC 60904-3); correction procedure (IEC 60891 — procedure 4 for c-Si with unknown temperature coefficients); qualification data (IEC 61215-2:2021, noting MQT 20–22).
2. Commissioning frame. Documentation and tests to IEC 62446-1.
3. IR survey. Preferably at irradiance >600 W/m², no clouds, low wind; artefacts excluded; findings classed against the TS 62446-3 abnormality matrix.
4. EL on the localised modules. Forward-biased, per TS 60904-13, in the lab or near-dark on site.
5. I-V trace. On a sample of strings; do not read differences within the instrument's ~5% accuracy as a module finding without an uncertainty budget; remember hot spots not yet costing power will not show.
6. String context. Reproduce string operating conditions before judging a single module's curve.
7. Acceptance criterion. Decide whether the contract criterion is crack class or power; the two can diverge.

Scope limits

This guide reads the IEC webstore abstracts of IEC 60904-3:2019, IEC 60891:2021, IEC 61215-2:2021, IEC TS 60904-13:2018, IEC 62446-1:2016 and IEC TS 62446-3:2017 (no clause texts), the IEA PVPS Task 13 review of IR and EL imaging (2018), the Task 13 degradation and failure report (2025) and NREL's O&M best-practices guide. It quotes no Chinese standard (GB/T 6495, GB/T 9535) or field-inspection rule, no module datasheet tolerance, no commercial inspection specification or price and no drone standard; the crack statistics are one batch at two sites. Not engineering, procurement or investment advice.