# The Field Number Is the Spec: 55 Plants, 19.92% EL Cracks

> A 55-plant survey measured EL cracks in 19.92% of modules and hot spots in 1.51%. The spread between plants, 350 times, is the number worth acting on.

![The Field Number Is the Spec: 55 Plants, 19.92% EL Cracks](https://cdn.ooitech.com/static/upload/image/20261026/field-el-crack-hot-spot-rates-55-plant-survey.png)

- ** Ooitech Editorial Team
- ** 2026-11-17
- ** 0 Views
- ** [Energy Storage & Applications](./Energy-Storage.html)

Editorial responsibility for this article rests with **Wuhan Ooi Photoelectric Technology Co., Ltd.** Written and reviewed by the Ooitech editorial team. See our [editorial policy](/editorial-policy.html) for how we write, where AI tools are used, and how to request a correction.

## The Field Number Is the Spec: 55 Plants, 19.92% EL Cracks

**Quick answer:** A 2026 conference survey sampled 55 Chinese plants across six regions and measured two failure modes separately: electroluminescence on 29 plants, drone infrared on 33. EL-visible cracks averaged 19.92% of modules per plant, hot spots averaged 1.51%. More than half the plants sat above 10% cracks and four in ten sat above 1% hot spots. The averages are the least useful part of it. The best plant measured 0.15% cracks, the worst 52.63%, a spread of roughly 350 times, and hot spots spread by more than 2,000 times. Plants shipping comparable modules land two orders of magnitude apart, which points at handling, mounting and measurement practice rather than at cell technology. What follows is what the survey measured, what an abstract cannot tell you, and how to turn the number into a defect budget per production stage.

Every module leaves the factory with an electroluminescence image and a pass stamp. Ten years later a drone finds cracks in one module out of five. The distance between those two facts is what this survey measures.

## What the survey actually measured

The team drew 55 plants at random across the northwest, northeast, north, east, south and central regions, covering different grid-connection years and different plant types. Two separate inspections ran side by side, which is the first thing to notice: cracks and hot spots were not measured on the same plants.

- **EL on 29 plants:** an electroluminescence image of modules, read for cracks inside the cells.
- **Drone infrared on 33 plants:** an aerial thermal survey, read for modules running hotter than their neighbours.

The comparison dimensions were grid-connection age, terrain, application scene such as agrivoltaics, and region. Cell technology was not broken out. The paper is a conference abstract, so the numbers below are the ones the abstract states, not the full dataset.

## The two averages, and why they are the wrong takeaway

Cracks averaged 19.92% of modules per plant. Hot spots averaged 1.51%. Read as headlines, the first number says one module in five has a visible crack and the second says hot spots are rare. Both readings are technically correct and practically useless, because neither describes the distribution.

The distribution is the finding. Cracks ranged from 0.15% at the cleanest plant to 52.63% at the worst. Hot spots ranged from 0.003% to 8.89%. Two plants can both report a passing factory audit and land three orders of magnitude apart on a field survey. An average across that population describes no plant in particular.

The threshold counts say the same thing more bluntly. 58.62% of the EL plants measured above 10% cracks, which is 17 of 29. 40.54% of the infrared plants measured above 1% hot spots. A number that half the population exceeds is a description of the population, not a warning about the outlier.

| Indicator | EL cracks | Drone infrared hot spots |
| --- | --- | --- |
| Plants inspected | 29 | 33 |
| Mean share of modules | 19.92% | 1.51% |
| Range | 0.15% to 52.63% | 0.003% to 8.89% |
| Plants above threshold | 58.62% above 10% | 40.54% above 1% |
| Ratio best to worst | about 350 times | more than 2,000 times |

## Four factors the survey names

The survey relates the two rates to four variables, and the pattern is consistent with how these defects form.

One caution before the list. These are associations measured on a sample of 55, not controlled comparisons. The same plant can be old, mountainous and dusty at once, and a survey of this shape cannot attribute a share of the 19.92% to any single factor. What it can do is tell you where to look first when your own number comes in high.

### Age

Both rates rise with grid-connection years. That fits the life of a crack: a microcrack starts at a busbar or a cell edge, extends under thermal and mechanical cycling, then the cell region it isolates stops contributing. A crack that has not isolated anything reads identically on an EL image, which is why age moves the visible count up faster than it moves the power.

### Terrain

Mountain plants measured highest for cracks, which the paper attributes to terrain and temperature swing. Transport and double handling are a plausible addition: a mountain site means a longer road, more transfers and a rack foundation that can twist a frame that was flat when it left the factory.

### Scene

Agrivoltaic plants measured highest for hot spots. Crop shading, higher humidity under the array and dust accumulation all push in the same direction, and shading is the trigger that turns an existing crack into a local heat source. The physics of how a shaded cell goes into reverse bias and heats is worth reading separately from the statistics; see [why shading and hot spots are not the same event](https://www.ooitech.com/does-shading-always-cause-solar-hot-spots-reverse-bias.html).

### Region

The northwest measured highest on both. Large diurnal swing, high irradiance, wind-blown sand and dust deposition are the standard explanation, and they act on cracks and hot spots through different paths: swing drives crack extension, soiling drives the shading that makes a cracked cell heat.

The two paths are not equally distributed across cell designs, either. A cell architecture that spreads reverse-bias current over a larger area runs cooler at the same shading ratio, which is one reason hot-spot counts cannot be compared across plants without also comparing what was installed. The comparison of [shading tolerance across cell architectures](https://www.ooitech.com/why-do-bc-solar-cells-offer-better-shading-tolerance-and-lower-hot-spot-temperature.html) covers that mechanism.

## Four things the abstract cannot tell you

The paper is honest about its own limits, and those limits are worth carrying into any decision.

- **Severity is not in the number.** A visible crack ranges from a hairline that never isolates a region to a fully disconnected cell area. The industry grades these A, B and C for a reason. A plant at 20% with mostly grade A is not the same plant as a plant at 8% with grade C.
- **Cell technology was not split out.** The abstract does not separate PERC, TOPCon and HJT, nor wafer thickness, busbar count, or glass-backsheet versus glass-glass. Those are precisely the variables a module maker would want next.
- **One percentage does not divide.** 58.62% is exactly 17 of 29, which matches the EL sample. 40.54% does not divide into 33 plants: 13 of 33 is 39.4% and 14 of 33 is 42.4%. It does divide into 15 of 37. The hot-spot denominator may not be 33, and until the full paper settles it, treat the hot-spot threshold share as approximate.
- **Age carries generation.** Older plants also run older modules built to older standards and installed to older practice. The time effect and the generation effect are mixed, and no field survey of this shape can separate them.

## Why a clean factory EL result does not predict this

A factory EL image is one measurement at one moment, taken before the module has been packed, transported, lifted, clamped and loaded. Everything that happens afterwards is invisible to it, and most of the crack population in the field is created there.

That is not an argument that the factory image is worthless. It is an argument that the image proves a narrower thing than buyers assume: it proves the module left the line without a flaw the line could detect, at the sensitivity the line uses. The field number is cumulative from the gate onward, so comparing it against a factory pass rate is comparing two different quantities.

Sensitivity deserves its own sentence, because it is where two plants with identical modules produce different percentages. An electroluminescence image taken at lower forward current, with a coarser camera or a shorter exposure, will not resolve the hairline that the same module shows under a slower capture. Nothing about the module changed. The reading did.

We run our electroluminescence inspections at sustained forward current with long exposure settings to detect hairline cracks early, ensuring fine defects are documented before modules ever leave the manufacturing facility.

That is also why the survey's 0.15% and 52.63% should be read as a statement about plants, not only about modules. A plant that reads its own images leniently will report a rate that looks excellent and will learn about the difference at the ten-year drone flight. Equipment and procedure for in-line and offline imaging are covered in [our note on EL and IV test equipment](https://www.ooitech.com/solar-panel-el-tester-and-vi-tester-machine-opt-m960b-m951b-m950b-ooitech-solar-module-el-testing-equipment.html), and the wider test menu in [the reliability test list](https://www.ooitech.com/pv-module-reliability-tests-list-cell-vs-module-failure-causes.html).

![field crack and hot spot rates](https://cdn.ooitech.com/static/upload/image/20261026/field-el-crack-hot-spot-rates-55-plant-survey.png)

## Where cracks enter after the gate

Almost every mechanism below is a known one, and each has a place it could be measured. The point of listing them is that the field number is the sum of five stages, and a plant that instruments four of them can attribute the fifth.

Stringer and layup set the initial population through soldering stress and ribbon dimples, which is a function of how the [ribbon is made and tinned](https://www.ooitech.com/how-pv-ribbon-is-made-drawing-annealing-and-tinning.html) as much as of the soldering station. Lamination adds edge stress and, at the wrong ramp, voids along the ribbon. Packing and transport add corner and stacking loads that a shock logger would record. Installation adds clamp torque, mount twist and the occasional foot. Service adds snow load, wind vibration and the thermal cycles that extend whatever is already there.

To prevent lamination from contributing to the 19.92% defect rate, we program multi-stage pressure cycles and uniform temperature ramps to avoid cell edge stress and eliminate voids along the ribbon.

Transport is the stage most often left out of the budget and the cheapest to instrument. A shock and tilt recorder in a shipping container costs less than a single EL re-read of the modules it protects, and it converts a vague suspicion into a date, a route and a peak acceleration. Sites that receive modules by long mountain road are exactly the ones the survey found at the top of the crack table.

## A defect budget you can write down

The practical output of a survey like this is not a target percentage. It is a split. Assign each stage a share, a measurement and an owner, and the field number becomes something a plant can argue about instead of something it can only report.

| Stage | Dominant mechanism | What to measure | Owner |
| --- | --- | --- | --- |
| Stringer and layup | Solder stress at the busbar, ribbon dimple | In-line EL after stringer, per shift | Line quality |
| Lamination | Edge stress, voids along ribbon | Post-lamination EL plus visual | Process engineering |
| Packing and transport | Corner impact, stacking load | Shock logger per shipment, arrival EL sample | Logistics |
| Installation | Clamp over-torque, mount twist, foot traffic | Torque audit, EL sample after install | EPC |
| Service | Snow, wind vibration, thermal cycling | Annual infrared plus string IV scan | O and M |

## What to measure, and how to grade it

Two rules make field numbers comparable. The first is to report the grade distribution, not the pass count. An EL image read as pass or fail throws away the only information that correlates with power loss, which is whether the crack isolates a region.

The second is to report the denominator with every percentage. This survey shows why: 58.62% resolves cleanly into 17 of 29, and 40.54% does not resolve into 33 at all. A field report that states the share without the count cannot be audited, and a number that cannot be audited will eventually be argued about instead of acted on.

For infrared, comparability depends on conditions. Apparent temperature difference moves with irradiance, wind speed and ambient temperature, so a survey flown in July and one flown in January are not the same measurement. Compare like with like, and tie the cadence to string-level underperformance rather than to the calendar.

A third rule is cheaper than either: keep the images. A crack count without the image behind it cannot be re-graded when a customer disputes the reading, and the cost of storage is trivial next to the cost of re-flying a plant.

## Where these numbers come from

Every figure in this article comes from a conference paper summarising a random survey of 55 Chinese plants, as reported at abstract level: 29 plants for electroluminescence, 33 for drone infrared, across six regions and a range of grid-connection years. The survey did not disclose cell technology or module bill of materials, which we note rather than fill in.

The arithmetic checks and the stage-level budget are ours, built from the mechanisms rather than from the survey data. The note about 40.54% not dividing into 33 plants is a check on the abstract, not a claim that the underlying work is wrong; it means the hot-spot sample may have been larger than the abstract states.

The grading convention, the comparability conditions for infrared, and the stage split are standard practice in field inspection. We have not attached a confidence interval to any figure, because an abstract does not carry one.

## FAQ

### Q: What does 19.92% refer to, modules or cells?

Modules. It is the average, across the 29 plants inspected by EL, of the share of modules per plant where a crack was visible in the image. It is not a cell-level rate and it is not a power loss figure.

### Q: Does a visible crack always reduce power?

No. It reduces power when the crack isolates a cell region from the current path. A hairline that crosses a busbar without disconnecting anything can sit in a module for years, and grading by severity is how the industry separates the two cases.

### Q: Why is the spread between plants so large?

The survey points at age, terrain, scene and region, and those are real. A 350-fold spread also needs handling and measurement practice to explain it: transport routes, mounting discipline and the sensitivity threshold used when reading the image all move the count.

### Q: Which cell technology has the highest crack rate?

The abstract does not say. It reports no breakdown by PERC, TOPCon or HJT, and no breakdown by wafer thickness, busbar count or glass configuration. Thin wafers and larger formats are generally more sensitive, but this survey does not quantify that.

### Q: Can hot spots be predicted from the crack data?

Only as a precondition. A crack creates the current path that can heat locally, and shading is what drives current through it. The survey measured the two on different plant samples, so it cannot give a conversion rate from cracks to hot spots.

### Q: How many modules should a site survey sample?

The abstract does not specify a sampling plan. What it does show is that the denominator has to travel with the result: 58.62% resolves to 17 of 29, and 40.54% does not resolve to 33. Publish the count.

### Q: Are drone infrared results comparable between seasons?

Not directly. Apparent temperature difference depends on irradiance, wind and ambient temperature, so seasonal comparison mixes the defect with the weather. Fly comparable conditions and track the same strings over time.

### Q: What should a module maker do with this number?

Split it across the five stages that can create or extend a crack, give each stage a measurement and an owner, and report EL by grade rather than by pass or fail. A single target percentage invites argument; a budget invites instrumentation.

## Final thoughts

The useful number in this survey is not 19.92%. It is 350, the ratio between the cleanest and the dirtiest plant in a random sample of 55. Modules of broadly similar construction do not produce that spread on their own. Handling, mounting and measurement discipline do, and those are things a plant can instrument, budget and audit. Start by reporting the count behind every percentage.

OOITech manufactures single and double chamber [solar panel laminator machines](https://www.ooitech.com/solar-panel-laminator-machine.html) with oil-heated and electric thermal configurations. We help module makers minimize mechanical stress and voids through precise thermal and vacuum profiling. Send your module stack, component drawings, or material samples to our engineering team for process evaluation and recipe testing.

---

##### Tags :

![](/template/ooitech/assets/img/shape/06.png)

![](https://cdn.ooitech.com/runtime/image/w800_h700_fitblur_v2_w800_h700_fitblur_v2_1757399770541443.webp)

### Request A Quote

All uploads are secure and confidential.

## We deliver expertise you can trust our service

Direct-from-Factory Equipment.

![](/template/ooitech/assets/img/icon/money-2.svg)

### Cost-Effective Advantages

We deliver exceptional value, maximizing results while optimizing budgets for clients.

![](/template/ooitech/assets/img/icon/staff.svg)

### Our Experience Team

Our skilled professionals specialize in innovative solutions and tailored strategies.

![](/template/ooitech/assets/img/icon/certified.svg)

### 15+ Years Industry Experience

Deep expertise ensures reliable, trend-aware, and proven outcomes for success.

![](https://cdn.ooitech.com/static/upload/image/20250910/1757477357667605.webp )

![](https://cdn.ooitech.com/static/upload/image/20250910/1757477724911512.webp)

![](/template/ooitech/assets/img/shape/06.png)

## What Our Client Say's about us

Client testimonials praise our deep understanding of their challenges, which leads to innovative solutions and strong ROI. Long-term collaborations—some over a decade—demonstrate their trust and satisfaction. Their success stories drive us to continually exceed expectations. [Know More **](#quote-form)

![](/template/ooitech/assets/img/icon/quote.svg)

Thanks to Ooitech for providing highly suitable BC solar cell experimental equipment.

![](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_1776426122864564.webp)

### KTECH

![](/template/ooitech/assets/img/icon/quote.svg)

A man with a big soul. Thank you very much for your visit Mr Wu. Thank you very much for the services provided in installing equipment and training my students. The kindest person and professional in his field

![](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_2026041716444445.webp)

### Jizzakh Polytechnic Institute

![](/template/ooitech/assets/img/icon/quote.svg)

thank you for ooitech's professional support and on time after service.

![](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_1784509010217979.webp)

### Diarra From Africa

![](/template/ooitech/assets/img/icon/quote.svg)

Thank you again so much again for the very big big help for improving and fixing the factory and also teaching the workers how to use the machines

![](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_1757479675272137.webp)

### Mark

BIPV Philippines

![](/template/ooitech/assets/img/icon/quote.svg)

Thanks to Ooitech for providing the fully automated production equipment—your installation and after-sales service have been excellent.

![](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_2026041720921238.webp)

### Amjad

## Our Latest Products

![Solar Panel Production Line: Configurations from 5 MW to 1 GW](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_2026032510951634.webp)

- [** ooitech](./solar-panel-production-line.html)
- [** 109](./solar-panel-production-line.html)

### Solar Panel Production Line: Configurations from 5 MW to 1 GW

Solar panel production line guide: stations, semi-automatic vs automatic, 5 MW to 1 GW configurations, layout, utilities, staffing and RFQ tips.

![CHT9951A Hipot Tester: PV Module Insulation 0.1-10kV](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_1774601338282260.webp)

- [** Rachael](./cht9951a-cht9951b-solar-panel-hipot-insulation-resistance-tester-pv-module-safety-testing-equipment.html)
- [** 58925](./cht9951a-cht9951b-solar-panel-hipot-insulation-resistance-tester-pv-module-safety-testing-equipment.html)

### CHT9951A Hipot Tester: PV Module Insulation 0.1-10kV

CHT9951A/CHT9951B hipot and insulation resistance tester for solar PV module testing. DC output up to 10kV, insulation resistance up to 99GΩ

![PV Ribbon Production Line: Drawing, Rolling, Tinning](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_2026051194879300.webp)

- [** ooitech](./pv-busbar-ribbon-integrated-drawing-rolling-tinning-production-line.html)
- [** 812](./pv-busbar-ribbon-integrated-drawing-rolling-tinning-production-line.html)

### PV Ribbon Production Line: Drawing, Rolling, Tinning

Professional integrated PV busbar ribbon production line combining wire drawing, rolling, flat drawing, annealing and tin coating processes for high

![SS-2500B Tabber and Stringer: 2400 Cells/h, 3-20BB](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_1774258018905303.webp)

- [** ooitech](./SS-2500B-Full-Automatic-Solar-Cell-Tabber-Stringer-Machine-High-Speed-Production-Line-Equipment.html)
- [** 37139](./SS-2500B-Full-Automatic-Solar-Cell-Tabber-Stringer-Machine-High-Speed-Production-Line-Equipment.html)

### SS-2500B Tabber and Stringer: 2400 Cells/h, 3-20BB

SS-2500B full automatic tabber stringer machine for crystalline silicon solar cells with 2400PCS/H capacity, featuring infrared soldering

![PV Ribbon Machine: Wire Drawing for Solar Ribbon](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_2026051187922761.webp)

- [** ooitech](./wire-drawing-machine-for-solar-ribbon-production-line.html)
- [** 780](./wire-drawing-machine-for-solar-ribbon-production-line.html)

### PV Ribbon Machine: Wire Drawing for Solar Ribbon

Professional intermediate wire drawing machine for solar ribbon production line, featuring four-axis horizontal design

![Full Automatic Solar Panel Production Line Equipment](https://cdn.ooitech.com/runtime/image/w800_h600_fitblur_v2_1774401482439949.webp)

- [** Rachael](./full-automatic-solar-panel-production-line-equipment-ooitech.html)
- [** 98980](./full-automatic-solar-panel-production-line-equipment-ooitech.html)

### Full Automatic Solar Panel Production Line Equipment

Ooitech full automatic solar panel production line covers glass loading, EVA laying, string layout, tape sticking, lamination, trimming, framing
