FA Lens Resolution: What the Figure Actually Means

Three different numbers are all called resolution

A fixed focal length lens datasheet can carry three unrelated quantities under the same word. The first is the pixel pitch the lens is claimed to support. The second is the modulation transfer at a stated spatial frequency. The third is the smallest object-space detail that still reaches the detector with usable contrast. They are not interchangeable, and a comparison built on mixed definitions will rank lenses in the wrong order.

The practical consequence is that two models quoted at 5 megapixels and 12 megapixels respectively may have identical object-space resolution at the working distance you intend to use. The megapixel figure describes a compatibility envelope, not a sharpness ranking. When a supplier quotes only that number, the specification is incomplete rather than strong.

The table below uses object-space resolution only, because that is the quantity that answers whether a defect of a given size will be visible. Every figure comes from the measured specification of an individual POMEAS FA lens model rather than from a family-level claim.

Resolution also has to be read together with the sensor. A lens that resolves finer than the sensor can sample produces no additional information; the sampling limit takes over and the extra optical performance is wasted. Matching the two is the point of the selection exercise, and it is covered separately in the pixel-pitch section below.

Finally, resolution is measured under stated conditions. Object distance, aperture setting and illumination wavelength all move the figure. A number quoted without those conditions cannot be reproduced, which is why the ranges given here are always tied to the model they came from.

Pixel pitch is a compatibility limit, not a sharpness figure

When a lens is described as suitable for a sensor with a given pixel pitch, the claim is that the lens will not be the limiting element on that sensor. It says the lens is good enough to avoid becoming the bottleneck. It does not say the resulting image will be sharp, and it certainly does not rank one lens above another that supports a smaller pitch.

This matters when the application is limited by something other than the lens. If diffraction, illumination contrast or the part itself sets the achievable detail, upgrading to a lens rated for a finer pixel pitch changes nothing in the image. The money is spent on a specification that the rest of the optical chain cannot use.

The reverse mistake is also common. A lens rated for a coarse pitch is used with a fine-pitch sensor, and the lens becomes the limit. The image looks soft at every aperture setting and no amount of processing recovers the detail, because it never reached the detector.

A reasonable procedure is to start from the smallest defect the inspection must catch, convert it to object-space resolution, and only then check which models meet that figure at the working distance available. Pixel pitch is a filter applied afterwards, to confirm the chosen model will not clip the sensor.

In the POMEAS FA range the measured object-space resolution spans 12 to 1000 um across eleven models that publish the figure, which is a wide enough spread that the choice of model matters more than the choice of sensor.

Object-space resolution moves with working distance

The same lens does not deliver the same object-space resolution at every focus position. Extending the working distance to make room for illumination or for a robot approach generally costs resolution, and shortening it recovers some of that back. A comparison made at each model's closest focus will not predict performance at the distance the machine actually uses.

This is the most frequent error in lens selection spreadsheets. Two models are compared using the numbers printed on their datasheets, which were measured at different working distances from each other. The comparison looks decisive and is meaningless, because neither number applies to the installation.

The correct procedure is to fix the working distance first, from the mechanical layout, and then compare models at that distance only. Models that cannot reach the required distance are removed from the list before any optical comparison begins.

When a supplier can provide the resolution at the application's working distance rather than only at the nominal one, that is a stronger signal than any headline figure. It means the measurement was made at a condition someone can use.

In the POMEAS FA range working distance spans 6.6 to 690 mm across the models where it is published, so the same optical family covers layouts from a compact inline station to a stand-off installation above a conveyor.

Distortion and resolution do not trade evenly

It is tempting to assume that the sharpest lens in a range is also the most geometrically accurate. On fixed focal length lenses that assumption fails often enough to be dangerous. Wide-angle designs carry more distortion than long focal length designs, and the model with the lowest distortion figure is frequently not the model with the best resolution.

Across the POMEAS FA range focal length spans 8 to 82 mm and TV distortion spans 0.002 to 3.73 %. The two extremes do not coincide, which means a selection that optimises one of the two figures will usually sit somewhere in the middle on the other.

For a measurement application the distortion figure has to be handled explicitly. Either the model is chosen so that distortion stays inside the tolerance without correction, or a calibration map is built and applied. The first route constrains the focal length; the second constrains the software.

For a presence or classification application distortion is usually irrelevant, and the resolution figure alone decides. Knowing which of the two the application needs is the step that prevents over-specifying the wrong parameter.

Where both matter, the practical answer is to shortlist by resolution at the application working distance, then check distortion for each shortlisted model before committing.

What to ask before accepting a resolution claim

Four questions resolve most ambiguity in a resolution claim. At what working distance was it measured. At what aperture. At what wavelength. And is the figure object-space or image-space. Without those four the number cannot be compared with anything else.

A supplier who can answer all four is describing a measurement. A supplier who answers none is repeating a marketing figure. The distinction is worth making early, because it determines how much of the selection work can be delegated.

The second useful question is what the figure refers to physically: a modulation value at a stated spatial frequency, a line-pair count, or a spot size. These produce different numbers for the same lens and are all legitimately called resolution.

For an application that must detect a defect of a known size, the most direct approach is to ask whether the lens resolves that size at the intended working distance, and to ask for the measurement condition that supports the answer. That converts an ambiguous specification into a testable claim.

POMEAS publishes object-space figures per model for the FA range, which is why the ranges in this article are given as intervals with a model count rather than as single headline numbers.

Where resolution stops being the deciding factor

Resolution is the deciding factor only while it is the coarsest limit in the chain. Once the lens resolves finer than the sensor samples, or finer than the illumination contrast allows, further optical performance adds nothing to the inspection result.

At that point the useful investment moves elsewhere: better contrast through illumination geometry, a more stable mounting, or a shorter exposure to reduce motion blur. Buying resolution beyond the sampling limit is the classic way to spend budget without changing the accept-reject outcome.

The check is straightforward. Convert the sensor pixel pitch to object space by dividing by magnification, and compare it with the lens resolution figure at the same working distance. Whichever is larger sets the achievable detail.

If the sensor is the limit, the next step is a sensor with smaller pixels, not a different lens. If the lens is the limit, the next step is a model with finer resolution at the same distance. Doing them in the wrong order wastes a purchase cycle.

This ordering also explains why resolution ranges are more useful than single figures when shortlisting: the range shows how much headroom exists before another element of the chain becomes the limit.

Shortlisting procedure for a fixed focal length lens

Start from the field of view and the working distance that the machine layout allows. These two, with the sensor format, fix the focal length. That step alone usually removes most of a catalogue from consideration.

Next, convert the smallest defect the inspection must detect into an object-space resolution requirement, and keep only the models that meet it at the required working distance. This is where a per-model measured figure replaces a family-level claim.

Then check distortion against the measurement tolerance if the application measures, or skip it if the application classifies. After that, confirm the mount and the image circle match the camera.

The remaining candidates are usually two or three models, and the final choice turns on availability, mechanical fit and price rather than on optics. That is the desired state: the optical decision should be closed before the commercial one begins.

Running the steps in this order means each figure is used only where it is decisive, which keeps the comparison honest and the specification defensible.

Frequently asked questions

Is a higher megapixel rating always better on an FA lens?

No. A megapixel rating describes the sensor the lens will not bottleneck, not the sharpness it delivers. Two lenses with different ratings can produce identical object-space resolution at the working distance you use. Compare object-space figures at the application distance before treating the rating as a ranking.

How do I convert lens resolution to the smallest detectable defect?

Take the object-space resolution figure at your working distance and compare it with the defect size. The defect must be at least as large as that figure to be resolved with usable contrast, and in practice two to three times larger for reliable detection. Anything finer than the figure is not recoverable by processing.

Why does my lens perform worse than its datasheet suggests?

Usually because the datasheet figure was measured at a different working distance or aperture. Extending the working distance to fit illumination or a robot approach reduces object-space resolution. Ask for the measurement condition and re-check at your own distance before concluding the lens is faulty.

Does distortion correction in software fix a poor resolution figure?

No, they are separate errors. Distortion moves image points to the wrong position and can be corrected with a calibration map. Resolution loss removes contrast and detail that never reached the sensor, and no calibration recovers it. A lens with good distortion figures can still be unusable for fine defect detection.

What aperture should I use for maximum resolution?

Most fixed focal length lenses peak two to three stops down from wide open, where aberrations have fallen but diffraction has not yet taken over. The exact setting is model specific, so it is worth measuring at your working distance rather than assuming. Wide open and fully stopped down are both usually worse.

Can one FA lens cover several inspection stations?

Only if the field, working distance and resolution requirement are compatible across all of them. In practice stations differ in at least one of these, and a lens optimised for one becomes a compromise elsewhere. Sharing a model across stations is worth doing when the optical requirements genuinely coincide, not as a procurement shortcut.

Lighting and camera lens selection

• Machine Vision Lighting Selection Basics
• Industrial Camera Lens Selection