1. Freeze the exact sensor configuration
A sensor family name or megapixel class is not enough. Record the exact part number, active array, optical format, pixel pitch, microlens architecture, color or RGB-IR pattern, cover glass and operating modes. Cropping, binning and HDR modes can change the useful image area or image-quality priority.
Treat a sensor change as an optical interface change. Two sensors with the same resolution may use different array dimensions, CRA acceptance and filter stacks.
- Exact sensor part number and revision
- Active width, height and diagonal
- Pixel pitch and readout mode
- CFA / RGB-IR, cover glass and filter stack
2. Translate pixels into useful optical frequencies
Pixel pitch sets the sensor sampling scale; the Nyquist frequency is one divided by twice the pixel pitch. It is a sampling boundary, not a universal lens pass/fail target. Select review frequencies from the smallest scene feature, algorithm need, demosaicing and signal-to-noise budget.
Evaluate more than one frequency. Lower frequencies describe broad contrast while higher frequencies describe fine detail, and both can influence perception quality.
- Pixel pitch and Nyquist frequency
- Task-relevant feature frequency
- Multiple MTF frequencies
- System MTF, noise and processing budget
3. Check image circle, field and CRA together
Image circle must cover the active sensor through tolerances, but coverage alone does not prove compatibility. At each field point, the chief ray reaches the sensor at an angle that interacts with microlenses, color shading and quantum efficiency.
Request sensor CRA guidance and compare it with the lens ray-angle curve across image height. Include lens decenter, sensor shift and assembly tilt rather than reviewing only nominal geometry.
- Active-array coverage with margin
- Lens CRA versus sensor acceptance
- Color shading and relative illumination
- Decenter, shift and tilt variation
4. Model the real sensor-filter-lens stack
Cover-glass and filter thickness, refractive index, wedge and distance to the sensor affect focus and aberration. Wide-angle rays also see a different effective spectral response through interference filters.
Use the intended visible, NIR or RGB-IR weighting and the actual f-number. Recheck focus and spectral behavior when the filter supplier, thickness or coating changes.
- Cover-glass and filter thickness
- Index, wedge and spacing
- Spectral weighting at field angle
- Visible/NIR focus relationship
5. Validate the complete camera, not a paper match
Nominal curves cannot capture every manufacturing and processing effect. Build representative camera modules and review full-field MTF, shading, color, noise, HDR artifacts and task images at relevant temperature and illumination conditions.
Freeze the sensor revision, lens revision, filter and ISP assumptions in the validation record. Define what must be repeated after an engineering change.
- Full-field camera MTF
- Shading, color and HDR scenes
- Temperature and unit-to-unit samples
- Configuration and change control