Electroluminescence Imaging

Every cell, every crack, in one picture.

Name: Avalon Electroluminescence Imaging
Brand: Avalon ST

A 65-megapixel electroluminescence picture of the whole panel, 0.2 mm per pixel, at two currents in one run. Then the anomaly analysis compares it with a reference of its panel type and tells you what is wrong, where, and how much it costs the cell.

Talk to an engineer See the analysis

Electroluminescence picture of a 144 half-cut cell module at 0.8 A in the fire colour map: a row of weak cells stands out in dark purple — A real result, not a rendering: 144 half-cut cell panel at a tenth of Isc, in the fire colour map. The dark cells are shunts.

65 MP

picture of the whole panel

11 232 x 5 804 px for a 2.3 m module

0.21 mm

per pixel

4.8 px/mm from corner to corner

5 MP

per camera

45 synchronised cameras on the bench

2 shots

per run, or up to 8

Isc and 0.1 x Isc by default

~3 s

anomaly analysis

per shot, cell by cell

One half-cut cell at native resolution with a curved crack and a T-shaped dark defect

Move over the cell, or touch it, to look closer.

Zoom into one cell. Then keep zooming.

This is one half-cut cell of the panel, cut straight out of the picture at its native resolution, no upscaling, shown with the viewer's default display sharpening. 462 x 900 pixels for a 91 x 182 mm piece of silicon, about 415 000 pixels. Every camera of the bench is a 5-megapixel sensor with wide-angle optics close to the panel, so the pixels go to the silicon, not to the distance.

4.8 px/mm: across the whole panel, from corner to corner
462 x 900 px: for one half-cell, about 880 x 880 px for a full M10 cell
0.21 mm: per pixel: a crack is seen, located and measured in millimetres
65 MP: opened instantly at any zoom, from the panel to the pixel

In this cell: a curved micro-crack running from the top edge, a T-shaped inactive area across a busbar, and small dark spots. Every busbar line is resolved.

Dark frames, subtracted

Every run also takes frames with the bias off and removes them. Stray light and sensor offset do not end up in your picture.

Exposure that fits the panel

Each panel type has its own exposure, found by the station itself with a few test shots. No saturated cells, no wasted dynamic range.

It tells you when a shot is wrong

Saturated pixels, a lead off the panel, a camera that did not answer: the shot is flagged with a sentence a technician understands, kept for diagnosis and never analysed as if it were good.

Two currents, two truths

High current shows cracks and inactive areas. Low current shows what high current hides: shunts, weak cells and cells that drop out. An electroluminescence recipe takes both in the same run, with the panel never moving, so the two pictures line up pixel for pixel.

The same part of the panel at 0.8 A: a row of cells is clearly darker — Drag the handle. At 8 A this row of cells looks healthy. At 0.8 A four of them lose 36 to 59 % of their light: shunts.

Part of the panel at 8 A: all cells look similar — Drag the handle. At 8 A this row of cells looks healthy. At 0.8 A four of them lose 36 to 59 % of their light: shunts.

Up to 8 shots per recipe

Current as a fraction of the panel's Isc or in amperes, voltage limit, exposure: one recipe per panel type, picked automatically with the DUT.

One bias window

Dark frames once, then each shot in turn, with waits that end as soon as the light is steady. The run is short and the panel stays put.

Delta images

Divide one shot by the other, or a panel before and after a stress test, cell by cell with levels matched, to see exactly what changed.

The anomaly analysis

Every picture is compared with a reference of its panel type, made from pictures of good panels. What is not as expected is painted on the picture, fine things in fire, broad areas in cyan, then sorted into findings and scored cell by cell. No threshold to tune per defect, no rule to write.

One number per cell

Each cell gets a quality index: 100 % is a perfect cell, 0 % no cell at all. A cell that is only a little darker as a whole loses nothing, a crack counts by its length, a dead area by its size. Sort by it, bin by it, stop the line on it.

Two shots know more than one

When a run has two shots, the software puts their analyses together.

findings confirmed

Found in both shots of the run, so they are not the noise of one frame. One switch hides the small marks that only one shot shows.

Shunt

darker at low current

A cell or a patch that is darker at 0.8 A than at 8 A is tagged as a shunt, with both percentages written next to it.

darker at high current

Darker at 8 A than at 0.8 A is a series resistance: a contact or interconnection problem, not the silicon.

Delta: what changed between two pictures

Divide one picture by the other and the software shows, pixel by pixel, where the panel behaves differently: the 8 A shot against the 0.8 A shot, or the same panel before and after a stress test. Both pictures are first brought to the same level, and where that level is taken decides what you see.

Delta of the 8 A and 0.8 A shots with levels matched over the whole panel: whole cells in orange or blue

The delta of the example run, 8 A divided by 0.8 A, computed from the two real pictures with the software's default settings.

Before and after a stress test

Any two results of the same panel can be compared: before and after damp heat, thermal cycling or a mechanical load. The pictures are lined up on the corner of the panel.

Four ways to match the levels

Each cell, around each point for the finest detail only, the whole panel, or as taken. Ratio or difference, smoothing, cell borders left out, range and threshold are all yours to set.

The cells that changed most

Next to the picture: a histogram of the delta, its mean and spread, the share of the panel beyond the threshold, and the twelve cells that changed most, one click to open each.

Cells worth a closer look

Four cells of the example, at native resolution. Hover a picture to take the map away and see what is under it.

B24 · 8 A

Quality index59.3 %

A strip broken off

A large part of the cell is cut off and dark: a broad area in cyan. The lowest cell of the panel at 8 A.

Dark patch
Spot
Dark cell

B22 · 8 A

Quality index67.4 %

Several cracks

The curved crack and the T-shaped one of the close-up above, traced in fire, with small spots around them.

Crack
Spot
Dark patch

C12 · 8 A

Quality index71.7 %

A crack and a degraded part

A long crack with a degraded part beside it, 30 % darker than the rest of the cell, as dark at both currents.

Degraded part
Spot
Crack

C23 · 0.8 A

Quality index27.4 %

A shunt

59 % darker than a typical cell at 0.8 A and only 8 % at 8 A: darker at low current, so the software calls it a shunt.

Degraded part
Crack
Dark cell

Hover to hide the map

A viewer built for metrologists

The electroluminescence workspace sits next to the IV results in PVStudio.

Deep zoom
Tiled image pyramid: the full 65 MP picture pans and zooms without waiting.
Cell loupe
Hover any cell to enlarge it, step from cell to cell with the keyboard.
Colour maps
Grey, inferno, magma, viridis, turbo and more, with levels and contrast equalisation: for the eye only, the data is never re-encoded.
Anomaly layers
Fine and broad layers, threshold, colours, and a key to hold the map away and see what is under a mark.
Visible-light overlay
Attach and align a normal photo of the panel to compare with the electroluminescence picture.
Export
PNG, TIFF or JPEG of any result at full resolution, for your reports and your customers.

Your operator has the last word

Kinds can be hidden, and operators add their own marks on the picture. The marks are saved with the result.

In the measurement file

The electroluminescence shots are taken in the same run as the IV curve and saved in the same measurement, so a module's power and its picture never get separated.

Better with every release

The analysis engines arrive as signed, encrypted packages through Pasan Cloud. Your station improves without a reinstall.

Works with our simulators

Electroluminescence Imaging runs in PVStudio beside the I-V measurement, on the same station and in the same measurement file.

Nexun PRO MaxA++ A+ A++

Nexun PROA++ A+ A++

Nexun ULTRAA+++ A++ A+++

Nexun Mobile LabA+ A+ A+

Built around the electroluminescence standards

Electroluminescence imaging of PV modules has its own IEC technical specification, and qualification labs use it before and after every stress test. Electroluminescence Imaging is made for both.

IEC TS 60904-13

Electroluminescence of photovoltaic modules: the module in forward bias in the dark, a camera that sees the near infrared, one picture at Isc and one at a lower current. Our electroluminescence bench takes exactly that pair, Isc and a tenth of Isc, in one run.

IEC 61215-2 and IEC 61730-2

Qualification and safety testing: thermal cycling, damp heat, mechanical load. Labs image a module before and after each sequence; the delta compares the two pictures cell by cell and shows what the stress did.

IEC 60904-1

The I-V curve of the same run, measured on the Nexun or HighLIGHT simulator, is saved in the same measurement as the electroluminescence pictures: power and picture of one module, never separated.

Guide: electroluminescence testing of PV modules (IEC TS 60904-13)

From the people who calibrate the world's flashers

Pasan has built the reference solar simulators for module makers and test labs since 1978, and Avalon ST has made full-LED simulators since 2018. Electroluminescence Imaging comes from the same engineers: the cameras, the bench electronics, the analysis and the software are designed in-house, and we measure on them every day.

Designed by the team behind PVStudio, the software that runs our Nexun and HighLIGHT systems
In-house analysis engines, updated over the air
Fitted to almost every Nexun vertical system we ship
Supported remotely by the engineers who wrote it

See your own panel in it

Send us a module, or ask for the full-resolution example, and judge the detail for yourself.

Talk to an engineer Browse the Nexun systems