Bright-Field vs. Dark-Field Optical Inspection
Bright-field inspection emphasizes reflectance and low-spatial-frequency variation; dark-field inspection emphasizes scattered light from edges, particles, scratches, and defects. The right architecture depends on defect physics, surface stack, feature direction, wavelength, polarization, numerical aperture, throughput, and false-positive tolerance.
Why this decision matters
This choice affects more than nominal optical performance. It changes package volume, tolerance sensitivity, supplier options, alignment effort, calibration, test equipment, production yield, and the evidence required before release. The correct answer therefore comes from the complete operating condition and acceptance method, not from a single catalog value.
Key engineering decisions
- Characterize defect scattering before choosing geometry.
- Set illumination and collection NA independently when needed.
- Design polarization and spectral control around the stack.
These decisions should be captured in a requirement or trade study before the team commits long-lead components. Where requirements conflict, rank the product priorities explicitly so optimization does not hide a business decision.
Specification checklist
- Defect types and sizes
- Surface and film stack
- Field and throughput
- Wavelength and polarization
- Signal-to-background target
Every value should state the condition where it applies and how it will be measured. A specification without a defined test condition is not yet an acceptance requirement.
Common failure mode
A visually attractive image is optimized rather than the separability of real defect and process variation, producing unstable classification.
The practical remedy is to compare the nominal model, tolerance prediction, mechanical interfaces, and measured configuration together. Treating the symptom as an isolated lens or component problem often produces another build with the same system-level limitation.
Verification approach
Measure signal distributions for representative good and defective samples across position, orientation, focus, process variation, and contamination.
Record the hardware revision, source or scene, wavelength, aperture, field point, focus or alignment state, environmental condition, processing, and measurement uncertainty. This makes the result useful for design iteration and supplier transfer rather than only for a one-time demonstration.
What to send PAO
Send good and defective samples, stack data, defect targets, throughput, field, current images, illumination geometry, and classification results.
PAO applies this framework through semiconductor-equipment optics, from requirements and architecture through detailed design, prototype evidence, and manufacturing transfer.