First Production Learnings and Quality Engineering

The line started running a week ago. First modules off the lamination press, through final inspection, into the storage racks. Most of the learning hasn't been about big things. It's been about small ones.

From the floor

The first surprise was cell sorting. We knew on paper that cells from the same batch vary by efficiency class, Isc, and a few other parameters. What we didn't fully appreciate was how much that variance shows up in module power output if you aren't careful about which cells go into which string. Two cells with a 0.2 percent efficiency difference don't sound like much. Put one in a string with nine others at higher efficiency and the lower one acts as a partial brake on the whole string. We tightened sorting tolerances in week two.

EL rejection rates in week one ran higher than expected. Most of it traced back to handling at the stringer. A pickup head that's a fraction too aggressive can crack a cell without leaving a visible mark. The cell still passes flash test. It still passes visual. It shows up as a fine line in EL, which a year or two later becomes a propagating crack and a dead string.

Incoming inspection caught two things in the first batch of materials we hadn't expected. One was an EVA roll with adhesion drifting out of spec at the edges. The other was a batch of junction boxes where roughly one in fifty had incomplete potting. Both would have made it into modules. The second one would have shown up as field failures five to seven years from now.

What you can see, and what you can't

There's a clean way to think about defects. Some you can see with your eyes. Scratches on glass. Frame damage. Bubbles in the encapsulant. A busbar sitting a millimetre off centre. These fail at visual inspection.

The harder problems are invisible. Micro cracks. Cell mismatch. Solder joints that look fine but won't survive thermal cycling. Hot spots that only appear under load. Water ingress paths through edge seals that won't open up for three monsoons.

A module that fails visual gets rejected before it ships. A module with invisible defects ships, passes flash test in the warehouse, gets installed, and degrades quietly for years. By the time it shows up in plant data, it's widespread, and the cause is long gone.

This is why EL testing matters. It catches most of the invisible category. Micro cracks, broken fingers, contamination, dead zones, soldering issues. We run EL twice in the line. Once after stringing, once after lamination. The first catches handling cracks before they get sealed in. The second catches lamination damage.

But EL is one window. It doesn't catch electrical mismatch between cells, which only flash testing reveals. It doesn't catch thermal anomalies under load, which need IR imaging on a powered module. It doesn't catch bond integrity, which needs pull tests. It doesn't catch edge seal weakness, which needs wet leakage and damp heat testing.

A factory that relies on EL alone has a blind spot the size of every defect EL cannot see.

One thing worth saying

Here's the real shift in thinking from these weeks. Inspection finds defects after they've happened. Quality engineering tries to make sure they don't happen.

If we're rejecting cells with handling cracks at EL, the answer isn't better EL. The answer is fixing whatever in the stringer caused the crack. Pickup head pressure. Vacuum settings. Conveyor speed. Operator training. Room humidity. Most defects you find at inspection have a process cause upstream.

A line that only inspects is fighting symptoms. A line with quality engineering is fighting causes. The first one keeps running and keeps producing rejects. The second one slowly stops needing to reject things.

That's the harder work. It's also the part you can't really see from outside the factory. It shows up in the field, thirty years later, in modules that hold their numbers.