Machine vision lenses for quality inspection

Start with your smallest defect size and work backward. Estimate the pixels needed to detect it; as one example, a high-contrast defect read by a thresholding algorithm might budget 5 pixels per defect, validated on real samples. That budget sets your sensor resolution requirement. Working distance is usually fixed by your physical setup, and it determines focal length through the rectilinear focal length equation. Aperture then sets your depth of field. Work through the specifications in this order: defect size, sensor resolution, working distance, focal length, aperture, sensor format.

Name the task first: barcode/Data Matrix reading, OCR/lot code reading, label placement verification, or print-defect inspection. Barcode reading needs low distortion and enough pixels per module; OCR needs tighter resolution and contrast uniformity. Most stations choose between a compact M12 lens for flat surfaces and a C-mount lens with an adjustable iris for curved containers with depth-of-field pressure. Telecentric optics are rarely the right answer for label inspection.

Start with the task. Contamination detection inside the seal lane needs contrast and spatial resolution, with lighting geometry usually doing more work than lens choice. Seal-width verification needs low distortion (commonly sub-0.5%) to keep apparent width consistent across the field. Seal-position and continuity checks need field coverage and resolution along the seal length. For most stations, choose a C-mount lens with an adjustable iris when packaging presentation is variable, and a compact M12 lens when the head must stay small and the surface is consistent.

Start with the inspection task type. For threshold-based pass/fail fill checks, moderate low distortion around 0.5% TV is typically sufficient and cost per lane matters. The CIL083 8mm M12 at $19 and -0.5% TV distortion handles these checks well on compact and multi-lane systems. For measurement-grade fill estimation where image height maps to physical fill volume, target under 0.2% distortion as a starting point and calibrate the height mapping. The CIL533 16mm C-mount at -0.1% distortion or the CIL553 16mm C-mount at -0.09% distortion are the safer choices when geometric accuracy is a real requirement.

Start with the inspection task. Date and lot code reading requires enough pixels per character and low distortion to keep character geometry consistent across the field. OCR and character verification is more demanding because the algorithm must distinguish individual strokes. Print-position and registration checks require low distortion across the full field; a lens with 2% barrel distortion introduces apparent position errors that can exceed registration tolerances. Print-defect detection is primarily a contrast and spatial resolution problem. C-mount with an adjustable iris is the practical default for most fixed-station area-scan applications; M12 works when the head must stay compact on a flat surface.

Only if you are measuring dimensions where perspective error is unacceptable, such as checking heights or diameters from above. For most surface defect detection, barcode reading, seal checks, and print verification, a standard entocentric C-mount or M12 lens works and costs significantly less. Telecentric lenses require precise working distance control and have a limited field of view. Commonlands does not currently sell telecentric lenses; reserve them for true metrology applications through a specialist vendor.

Yes, for embedded inline inspection where size and cost are the primary constraints and the inspected surface is flat or near-flat. Most M12 lens models cover sensors up to 1/1.8", and select designs reach 2/3"; the low-distortion CIL052, for example, covers sensors up to 1/1.8". For larger sensors, precision measurement requiring an adjustable aperture, or depth-of-field control on curved or uneven surfaces, C-mount is the better tool: its iris ring makes stopping down for DOF practical when illumination is programmatically controlled.

Working distance sets focal length through the far-conjugate approximation EFL = (working distance × sensor width) / field-of-view width, which is close when the working distance is long relative to the focal length; exact thin-lens conjugates give a slightly shorter figure. Shorter working distance requires a shorter focal length at a given field of view and sensor size. Longer working distance requires a longer focal length but provides more physical clearance for conveyor hardware, robot arms, or enclosures. Working distance is usually constrained by your physical setup before you ever open a lens catalog.

For flat parts at a fixed working distance, open the aperture for maximum light throughput, MTF permitting; many industrial lenses are aberration-limited wide open and reach peak sharpness one to three stops down. For 3D parts or components with varying heights, stop down to increase depth of field.

Two diffraction limits apply, both at 550nm. The depth-of-field crossover sits near F# of 2 times the pixel pitch in microns, roughly F/7 on 3.45 micron pixels, and marks where softening outweighs the depth you gain. The tighter detection marker, where the Airy disk spans two pixels, sits near 1.5 times the pitch, or F/5.1 on the same sensor.

Yes, for most inline barcode and 2D code reading. A lens with significant barrel or pincushion distortion compresses or stretches code module geometry near the image periphery. Decoders read bar-to-space ratios and cell dimensions; when those are skewed by distortion, decode confidence drops and partial reads become more likely near the corners. Our own starting point is 0.3% or less distortion, which keeps symbol geometry close to nominal across the full field; the figure a given station needs depends on module size, where the code sits in the field, the decoder, and print quality.

Match your lens to your sensor's pixel pitch. A megapixel rating names the sensor class a lens targets; it is not proof of resolved pixel pitch. The 12MP-rated CIL512 targets 3.45 micron pixels on a 1.1" sensor, and the 25MP-rated CIL553 targets 25MP sensors with 2.74 micron pixels; confirm the pairing against measured MTF data at your pixel pitch, aperture, and field position. If the lens under-resolves the sensor, you are paying for pixels you cannot use: the optical blur exceeds the pixel pitch before the sensor limit matters.