One of the biggest mistakes I made early in my career was comparing optics by magnification alone.
What actually matters
- Resolution
- Working distance
- Depth of field
- Illumination compatibility
- Throughput
- Mechanical constraints
The math that replaces "magnification"
Two numbers tell you more than magnification ever will: sampling and numerical aperture.
Sampling - how big is one pixel on the object?
pixel on object = sensor pixel size / magnification
example: 3.45 um pixels at 2x -> 1.7 um per pixel
a 5 um defect spans ~3 pixels -> detectable, barely measurable
My rule of thumb: 3 pixels across a feature to detect it, 10 to measure it with confidence.
Numerical aperture - can the lens physically resolve the feature, and what does that cost in depth of field?
resolution limit ~ 0.61 * wavelength / NA
depth of field ~ wavelength / NA^2
NA 0.14: resolution ~2.4 um, DOF ~28 um
NA 0.28: resolution ~1.2 um, DOF ~7 um
NA 0.42: resolution ~0.8 um, DOF ~3 um
This table is the whole tragedy of high-resolution imaging in two columns: every step toward finer resolution costs you a square step in depth of field. A lens that resolves your defect but has less depth of field than your part warpage has not solved your problem.
My evaluation process
- Determine the smallest feature that must be resolved.
- Determine the required field of view.
- Determine the cycle time.
- Check the candidates on paper first: sampling, NA, depth of field versus part height variation, working distance versus mechanical reality.
- Test the survivors against real parts.
- Validate with production-representative samples, not just the nice ones.
Step 4 eliminates most candidates for free. I keep a one-page calc sheet per application and fill it in before requesting a single demo lens.
Things datasheets do not tell you
- Edge-of-field performance. Many lenses are quoted at center; corners can be 2x worse. Test with a grid target across the full field.
- Performance at your working distance, not the designed one.
- How the lens behaves with your lighting - telecentric lenses and coaxial illumination interact in ways the datasheet never mentions.
- Thermal drift of focus. A lens that holds focus from a 20 degree morning to a 35 degree afternoon is worth real money.
The lesson
The best optic is not the most expensive optic. It is the optic that solves the problem while satisfying all the constraints.