EL Testing Solar Panels: What Buyers Should Ask For and Read
EL testing shows cracks a flash test hides. The resolution, injection current, pass criteria and sampling to require — plus a pass/fail table.
EL (electroluminescence) testing passes a small forward current through a finished module and photographs its near-infrared glow, exposing cracked, mismatched or inactive cells that a flash test averages away. For a bulk order, ask for module-level imaging at 1 mm per pixel or better, injected at short-circuit current, with written pass criteria — zero cracks crossing a busbar, no dark area above 10% of any cell — and either 100% coverage or an ISO 2859-1 sample at AQL 1.0.
What EL Testing Reveals That a Flash Test Hides
An EL image shows where current is not flowing inside the module, which is exactly the information a flash test compresses into one watt number.
A flash test measures peak power at 1000 W/m², STC, and it repeats only to about ±3%. On a 550 W module that is a ±16 W band, so a hairline crack costing 8 W disappears inside the spread. EL does not measure power at all — it maps it. A crack crossing a busbar, a solder joint that never reflowed, a cell damaged at the stringer: each has a distinct signature, and each can pass a flash test on the day of production and fail in year three on a rooftop.
That is why EL sits inside IEC 61215 pre-conditioning and why buyers treat it as a proxy for handling discipline. A factory that images every module and keeps the files has a reason to be careful after lamination; a factory that images nothing has none.
Camera Setup: What "Good" Resolution Actually Means
A usable module-level EL image needs a near-infrared camera (silicon CCD or CMOS, sensitive roughly 900–1150 nm), at least 1 mm per pixel across the full module, a dark enclosure, and a stable DC supply injecting current at or near short-circuit current.
Resolution, focus and exposure
A 2.3 m module imaged at 600 px across gives about 4 mm per pixel — a hairline crack is one or two pixels wide and may vanish entirely. At 2000 px across you get roughly 1 mm per pixel, the workable floor for judging a crack; 0.5 mm per pixel is what good line-scan systems reach. Ask for the pixel dimensions of the delivered image, not the camera's megapixel rating. Exposure at Isc typically runs 5–60 seconds for silicon sensors; shorter exposures need InGaAs or heavier current, and noise rises accordingly.
Injection current and module temperature
IEC TS 60904-13 describes EL imaging of PV modules and is the reference to write into a contract. Require images at Isc (±2%) with the module at 25 ± 5 °C. Low injection current makes weak shunts look worse and healthy cells dimmer, so a report taken at 0.1 Isc cannot be compared with one taken at full Isc. Temperature matters too: a hot module shifts the emission and can hide a shunted region. If the supplier cannot state injection current and temperature, the images are marketing, not evidence.
How to Read an Electroluminescence Image
Bright means live silicon, dark means current is being lost — the skill is telling a harmless dark line from a serious one.
Dark fingers, busbars and solder joints
Short parallel dark lines are broken fingers. They are common, mostly cosmetic, and usually cost well under 1% of cell current. A dark band running the full width of a cell is a different animal: a broken or lifted busbar, which removes a third of that cell's collection and creates a local hot spot under reverse bias. Look at the soldered joints at the cell edges — a bright ring around a solder pad with a dark centre usually means poor wetting, and it will fatigue faster under thermal cycling.
Cracks, inactive areas and PID signatures
A crack appears as a thin dark line, often branching, and its severity depends on what it crosses. A crack inside one cell that stops short of the busbar typically costs 1–3 W on a full module; a crack severing the cell into two independently operating halves costs 5–15 W and worsens with thermal cycling. Fully dark cells or large dark blocks are rejects. Halo or ring patterns around cell edges, especially after damp heat, point to shunting of the kind covered in our PID explainer; modules tested to IEC 62804 should show no such signature.
EL Pass Criteria: What a Cell May and May Not Have
A defensible criterion names the defect, the location and the limit — "no crack crossing a busbar" is testable, "no serious cracks" is not.
| Defect | EL appearance | Typical acceptance | Action |
|---|---|---|---|
| Crack crossing a busbar or ribbon | Dark line across the full cell width | 0 per module | Reject and replace |
| In-cell crack not crossing a busbar | Thin branching line under 5% of cell area | Up to 2 cells per module | Accept with note on file |
| Broken finger | Short parallel dark lines | Up to 5 per cell if power loss under 1% | Accept |
| Cold or lifted solder joint | Dark spot with bright ring at the pad | 0 | Reject |
| Chipped corner | Black corner area | Under 2 mm and clear of the junction box | Accept; larger rejects |
| Inactive cell or block | Fully dark cell | 0 | Reject |
| Halo or ring shunting | Dark ring around the cell edge | None after IEC 62804 | Reject lot, investigate PID |
Most of these criteria are geometric, not electrical, and that is deliberate: an inspector can apply them in seconds on the line, and you can argue the point later with an image rather than a power measurement.
Full-Lot EL or an AQL Sample?
Image 100% of modules when the order is a full container or more and the end use carries a 25-year performance warranty; an AQL 1.0–2.5 sample is a reasonable compromise for small or repeat orders.
| Approach | When it runs | Coverage | Catches | Best for |
|---|---|---|---|---|
| String-level EL | After soldering, before layup | 100% on automated lines | Cracked cells, broken fingers, bad solder | Cell and string quality |
| Module EL after lamination | End of line, before packing | 100% or AQL 1.0 | Handling and lamination cracks | Serial-level warranty proof |
| AQL sample at pre-shipment | After packing, before loading | ISO 2859-1 level II, AQL 1.0–2.5 | Lot-level drift | Small or repeat orders |
| Post-freight EL at destination | After unloading | 1–3 pallets or an AQL sample | Shipping microcracks | Claim and insurance evidence |
| Witnessed EL run | During production | One day of output | Operator practice | First order with a new factory |
What a sample can and cannot prove
An AQL sample proves the lot is not systematically broken. It cannot prove any individual serial number is crack-free, and it will miss a defect that clusters — damage from one bad handling shift, or one dropped pallet in the warehouse. If your contract carries liquidated damages, sample-level evidence will not support the claim. Cost is rarely the blocker: module-level EL usually adds well under 1% to the FOB price, and off-line imaging of a finished pallet runs a few dollars per module, depending on the line and whether imaging is inline or manual.
Putting EL Criteria Into the Purchase Order
Write EL into the PO as a deliverable with a format, a deadline and a remedy, not as a promise in an email.
- Method: IEC TS 60904-13, injection at Isc ±2%, module at 25 ± 5 °C, dark-room conditions stated in the report.
- Image spec: minimum 1 mm per pixel over the full module, files named by module serial number and kept in raw format.
- Coverage: 100% above one 40ft HC container; below that, ISO 2859-1 level II with AQL 1.0 for critical and 2.5 for major defects.
- Timing: images of the finished, laminated module after cleaning — not the stringer — delivered before the shipment leaves under your chosen Incoterm.
- Retention: raw files held 24 months and re-issuable against a serial number on request.
- Remedy: modules failing the agreed criteria replaced free at the factory, with third-party retest cost carried by the failing party.
One clause is worth the argument: allow a third-party inspector to witness the EL run. SGS, TÜV or BV can attend for a day, and their presence changes how the operator sets the camera. It also gives you a baseline to compare against post-freight imaging, which is where most microcrack disputes actually start — our guide to microcracks after ocean freight covers that sequence step by step.
Inside the Factory: What I Actually Look At
On our line EL sits at two points — after stringing, where an operator sees a cell-level image within seconds, and after lamination, where the finished module is imaged at full resolution and the serial number is bound to the file.
What I look for in incoming cell inspection is not cracks, because the cell maker's own EL already handles those, but consistency: cells from different bins with visibly different brightness in the same string will mismatch under real irradiance even though every cell passed. That shows up as a patchwork module image long before it shows in a flash test. On finished modules the failures I see most are handling cracks near the junction-box corner and ribbon-solder shadows at the busbar ends, both of which appear in the first couple of percent of a lot if the line is drifting. Buyers who run a proper pre-shipment inspection catch them; buyers who accept a PDF of twelve marketing images do not.
Pairing EL with the other tests
EL is a structural check, so pair it with a flash test for binning, an insulation test to IEC 61730, and a PID screen if the destination is hot and humid. Check that the two data sets agree — a module showing 3% dark area but a clean flash result usually means an uncalibrated tester or a quiet re-bin. Confirming cell technology and verifying power-tolerance sorting close the remaining gaps. If you want to see the images we can produce against your own specification, send the criteria to our sales team and we will return a sample report for your BOM.
Frequently Asked Questions
Is EL testing worth the extra cost on a bulk solar panel order?
Usually yes. Module-level EL typically adds well under 1% to the FOB price, while a single container of modules with busbar-crossing cracks can cost you a warranty claim worth many times that. For orders under a few hundred pieces, ask for an AQL 1.0–2.5 sample instead of 100% imaging; the cost drops but you still get lot-level evidence.
What resolution should I require for an EL test report?
Require at least 1 mm per pixel across the full module, which means roughly 2000 pixels across a 2.3 m panel. Anything near 600 pixels across gives 4 mm per pixel and a hairline crack may not appear at all. Ask for the pixel dimensions of the delivered file, not the camera's megapixel rating, and insist files are named by module serial number.
How do I tell a serious crack from a harmless one on an EL image?
Look at what the dark line crosses. A crack confined inside one cell that stops short of a busbar typically costs 1–3 W on a full module and is usually acceptable. A crack that crosses a busbar, or splits a cell into two independently operating halves, costs 5–15 W and gets worse under thermal cycling, so it should be an automatic reject.
Should I require EL on 100% of modules or an AQL sample?
Use 100% module-level imaging for full-container orders and any project with a long performance warranty, because only serial-level images support a later claim. An ISO 2859-1 level II sample at AQL 1.0 for critical defects is defensible for small trial orders or repeat suppliers with a stable history, but it cannot prove a specific serial number is crack-free.
Can a module pass EL testing and still fail in the field?
Yes. EL confirms structure, not long-term stability. A module can image cleanly and still suffer PID in a hot, humid site, degrade through solder-joint fatigue, or arrive cracked after ocean freight. That is why EL is paired with a flash test, an IEC 61730 insulation test, an IEC 62804 PID screen where relevant, and post-freight imaging at destination.