Two datasheets can show the same focal length and the same F-number and describe lenses that perform completely differently. The difference lives in three lines most buyers skim past: MTF, distortion and relative illumination. This article explains what each number means, what a good value looks like, and how to compare two lenses honestly.
MTF — Modulation Transfer Function
MTF measures how much contrast a lens preserves at a given spatial frequency. Shoot a black-and-white bar pattern: at low frequency (wide bars) a decent lens keeps nearly all the contrast; as the bars get finer, contrast falls. Plot that fall-off and you have the MTF curve.
A datasheet usually quotes MTF at one or two frequencies in line pairs per millimetre (lp/mm):
| Class | Typical frequency | What “good” looks like |
|---|---|---|
| General FA inspection | 30–60 lp/mm | ≥0.4 at 30 lp/mm across the field |
| High-resolution FA / 20MP | 60–100 lp/mm | ≥0.3 at 60 lp/mm |
| Line scan (4K–8K) | 60–110 lp/mm | ≥0.3 at the specified frequency, centre and edge |
| 16K large format | 80–120 lp/mm | edge MTF within 20% of centre |
Read the edge value, not the centre. Almost every lens looks excellent on axis. The question is what happens at 80–100% of the image height, and that is where a cheap design loses 40% of its contrast while a corrected design holds.
MTF vs. Sensor Pixel Size — the Nyquist Link
There is no point buying more resolution than the camera can sample. The sensor's Nyquist frequency is:
A 3.1µm pixel has a Nyquist limit around 161 lp/mm; a 7µm line scan pixel around 71 lp/mm. Specify a lens whose MTF is still healthy near your Nyquist frequency — but do not pay for 200 lp/mm performance on a 7µm sensor.
Distortion — TV Distortion vs. Optical Distortion
TV distortion measures how much a straight line at the edge of the field bows, expressed as a percentage of the image height. It matters for two reasons: metrology (a bowed line measures wrong) and multi-camera stitching (adjacent fields of view will not align).
- <0.1% — metrology grade; required for gauging and dimensioning.
- 0.1–0.5% — fine for presence / absence, code reading and most defect detection.
- >1% — correctable in software, but budget calibration time and accept a small accuracy penalty.
Our line scan and macro families are corrected to well under 0.1% in the ranges where measurement matters, and every datasheet states the test conditions.
Relative Illumination (RI)
RI is corner brightness as a percentage of centre brightness. Below about 60% you will see visible shading that software must flatten — and flattening amplifies noise in the corners. Large-format and wide-angle designs are the usual offenders; a well-corrected industrial lens holds 70–90%.
What We Actually Test
Every CLARIOPTICS lens family is characterised in our own CMA / CNAS accredited optical testing center — interferometers for wavefront (RMS ≤ 1/200λ @ 632.8nm), MTF stations for frequency response, autocollimators for centration, and environmental chambers for thermal behaviour. Test data can ship with the lens on request, so the numbers in the datasheet are measured, not modelled.
A 60-Second Comparison Checklist
- Is the image circle ≥ your sensor diagonal? (no vignetting)
- Is MTF quoted at a frequency near your Nyquist limit? (resolution you can actually use)
- Is the edge MTF within ~20% of centre? (uniformity)
- Is distortion low enough for your measurement tolerance?
- Is relative illumination ≥ 70%? (no heavy shading correction)
- Does the datasheet state WD range and O/I conjugate? (fits your machine)
Need the Test Data?
Download the datasheet for any model from its product page, or ask our engineers for the full test report with your sample — we reply within one business day.