IN Brief:
- In-Sight 1750 combines AI-assisted reading, imaging, and dedicated illumination for semiconductor wafer and panel identification.
- The system targets difficult substrates and degraded markings that can otherwise interrupt automated production.
- Existing In-Sight 1740 jobs can be migrated to the new platform, which also adds browser-based setup and read logging.
Cognex has introduced the In-Sight 1750 Series for identifying semiconductor wafers and panels, combining AI-assisted reading with dedicated imaging and illumination hardware to improve recognition where markings, materials, or process conditions make conventional reads unreliable.
Identification follows material through fabrication, packaging, and test by linking a physical wafer or panel to manufacturing records. Serial numbers, characters, and machine-readable codes allow production equipment and factory software to associate each item with the correct process history, recipes, test information, and quality data.
The task becomes less straightforward after repeated processing. Semiconductor substrates can be reflective, patterned, coated, contaminated, or marked with characters whose contrast has deteriorated. Packaging processes add further variation through different materials and larger panel formats. A failed read can force an operator to intervene or a machine to repeat a step, reducing utilisation on equipment that is otherwise highly automated.
Cognex says the In-Sight 1750 combines its latest AI technology with advanced imaging and purpose-built illumination to improve performance across those conditions. The launch material does not provide a common recognition benchmark, so the practical improvement will depend on the actual substrate, marking method, optics, process residue, and alignment conditions found in each installation.
The system is also intended as an upgrade path from the In-Sight 1740. Existing jobs can be converted to the new platform, reducing the amount of engineering needed when a semiconductor equipment manufacturer or fab replaces an installed reader. That matters because identification hardware normally sits inside a larger machine whose mechanics, software interfaces, and process qualification may already be fixed.
A browser-based setup environment is intended to simplify configuration, while built-in read logging gives engineers a record of identification performance. Logging becomes useful when failures are intermittent because the same missed code can result from poor marking, contamination, changing reflectivity, mechanical positioning, illumination, or the reader configuration itself.
Those records can also support process control. A gradual increase in failed reads may indicate that marking quality has changed upstream or that an optical component is becoming contaminated. Treating the reader as a data source rather than a simple pass-or-fail sensor gives maintenance and process engineers more information when deciding whether the fault belongs to the identification system or the manufacturing step that produced the mark.
Traceability requirements increase as semiconductor assembly becomes more fragmented. A device may pass through wafer fabrication, probing, thinning, dicing, advanced packaging, final test, and multiple subcontractors before shipment. Maintaining identity through those stages is essential when a yield excursion or reliability failure has to be traced back to a particular lot, process step, or supplier.
Panel-level packaging broadens the problem further because the geometry and material stack differ from a conventional silicon wafer. Identification systems may need to read marks across organic or composite substrates, different surface finishes, and varying panel dimensions while remaining compatible with automated handling equipment. A single platform able to cover both wafers and panels can reduce the number of reader architectures maintained across a production line.
A successful optical read still forms only one part of the traceability chain. The identifier has to be transferred correctly to tool control or manufacturing execution software and matched with the correct process record. The automation also needs a defined response when read confidence falls below an acceptable level, otherwise better recognition hardware can still feed incorrect identity data into the factory system.
The In-Sight 1750 arrives as wafer fabrication and advanced packaging capacity continue to expand around AI and high-performance computing. More capacity means more identification points, while higher equipment utilisation makes unplanned intervention more expensive. That gives a specialised reader a measurable target: fewer failed reads and fewer production interruptions rather than a general claim that AI improves machine vision.
Cognex has not disclosed pricing or a complete optical specification in the launch material. Equipment builders will therefore judge the 1750 against the existing 1740 and competing readers on difficult-mark performance, integration effort, illumination flexibility, logging, maintenance, and the amount of operator intervention removed from sustained production.
