When a vision system measures badly, the camera gets blamed first. In our experience the limiting factor is usually the lens: the sensor can only record the image the lens delivers, and three lens-level quantities decide whether the numbers on your HMI are trustworthy.
1 — Distortion Sets Your Geometric Floor
Distortion is a geometric error: the lens images a straight line as a curve. If a datasheet promises ±0.1% TV distortion over a 100mm field of view, the worst-case geometric error at the field edge is 0.1% × 50mm = 50µm — before any software calibration. Calibration reduces the residual, but only where you placed calibration points; uncorrected local distortion still adds error between them.
Rule of thumb: match the distortion class to the tolerance you must hold. ±3% FA glass is fine for presence checks, ±0.1% for dimensional QA, and sub-0.05% or telecentric-class designs for metrology.
2 — Lens MTF Must Exceed Your Nyquist Frequency
A 5µm-pixel sensor has a Nyquist limit of 100 lp/mm. If the lens delivers 30% MTF at Nyquist, one-pixel features survive; at 0% MTF they vanish even though the camera “has 5-micron pixels”. Ask for MTF at your Nyquist frequency, at the field position you actually measure (edge, not only centre), at your working aperture — that is exactly what our datasheets tabulate.
3 — Relative Illumination Moves Your Noise Floor
Relative illumination (RI) is how much dimmer the field corners are than the centre. Low RI forces you to raise camera gain, which raises noise exactly where your defect threshold is tightest. For measurement, unit-to-unit consistency of RI matters as much as the absolute value: a spread between two lenses changes your grey-level thresholds at every batch changeover.
4 — Aperture: Depth of Field vs Diffraction
Stopping down buys depth of field but costs resolution: the diffraction cut-off of green light is about 320 lp/mm at F5.6 and about 230 lp/mm at F8. Wide open you gain contrast but lose depth across web flutter; over-stopped you soften pixel-level detail. For measurement stations we usually recommend F4–F5.6 and a mechanical fixture that keeps the standoff constant.
A 60-Second Audit of Your Current Station
- Compute Nyquist from your pixel pitch; check the lens MTF there, at field edge.
- Convert the datasheet TV distortion to µm at your field of view; compare against tolerance ÷ 5.
- Check relative illumination — above ~70% for measurement, or verify your threshold margin.
- Confirm the working-distance range covers your mechanical envelope with margin.
Related reading: how to read MTF, resolution and distortion on a datasheet. To shortlist lenses that meet a distortion/MTF budget, see the FA fixed focal series and macro & large format tables — or filter all models in the lens selector.
Still Unsure?
Send us the camera model, field of view and the tolerance you must hold — our engineers return a measurement-suitable shortlist within one business day. Start here →