The Biggest Mistakes Companies Make When Specifying Machine Vision Lenses

The lens is the most underrated component in a machine vision system. People will spend weeks choosing a camera, comparing sensors and resolutions down to the last pixel, then reach for whatever lens happens to be on the shelf to go in front of it. It is exactly the wrong way round. The camera can only capture the image the lens delivers to it. A poor lens choice caps the performance of everything behind it, no matter how good the camera is.

The frustrating thing is that lens mistakes are remarkably predictable. After years of specifying these systems, we see the same handful of errors again and again. Every one of them is avoidable with a few minutes of calculation before anyone places an order. This guide walks through the biggest mistakes companies make when specifying a lens, why each one hurts and how to get it right. There is a link to our lens calculator at the end so you can put the numbers into practice on your own application.

Mistake one: guessing the working distance

The working distance is the gap between the lens and the part you are inspecting. It is the mistake we see most often. People estimate it from where the camera looks like it will sit in the machine, order a lens, then discover on the bench that the lens will not focus at that distance, or that it does not give the field of view they assumed it would.

The reason this catches people out is that working distance, focal length and field of view are locked together. You cannot choose them independently. For a given lens and sensor, fixing the working distance fixes the field of view. If you need a particular field of view at a particular distance, that dictates the focal length of the lens. Guess one and get the others wrong. The fix is simple arithmetic done in advance. Measure the space you actually have, decide the field of view you need to see the part properly, then calculate the focal length that gives you that field of view at that distance. This is precisely the sort of thing a lens calculator does in seconds. It turns a common and expensive mistake into a solved problem before you buy anything.

Mistake two: ignoring sensor size matching

This one is less obvious and just as damaging. Every lens projects a circular image. That image circle has a size. Every camera sensor has a physical size too, described as a format like 1/1.8 inch, 2/3 inch, 1 inch or 1.1 inch. For the system to work, the lens has to project an image circle at least as large as the sensor it is sitting in front of. Put a lens designed for a small sensor on a camera with a larger one and the image circle does not cover the whole sensor. The result is dark, blurred, distorted corners, a problem called vignetting, where the edges of your image fall away into shadow and softness.

This matters more now than it used to, since sensors keep getting larger. Modern high-resolution cameras use large sensors. They need lenses designed to cover them. This is exactly why the lens ranges we supply specify the sensor format and resolution they are built for. A lens like the Computar MPX is designed to cover sensors up to a stated size. The larger-format Kowa and Computar high-resolution lenses are made for the big sensors in the newest cameras. The rule is straightforward. Check the sensor format of your camera, then choose a lens whose image circle covers it. A lens that is too small for your sensor cannot be fixed in software.

Mistake three: under-specifying resolution

People tend to think of resolution as purely a camera property, measured in megapixels. It is not. The lens has a resolution too. If the lens cannot resolve the fine detail, it does not matter how many megapixels the camera has, since the detail never reaches the sensor. A high-megapixel camera behind a low-resolution lens is a fast way to waste money, since you have paid for a sensor that can see detail the lens cannot deliver to it.

Lens resolution is often expressed as a megapixel rating, the sensor resolution the lens is designed to do justice to, or in line pairs per millimetre, a measure of the finest detail it can distinguish. The principle is the same either way. The lens has to be able to resolve at least as much detail as the camera can capture, or the camera is held back. This is why our higher-resolution lenses carry explicit ratings. The Computar MPY series is rated for 24 megapixel sensors and the Kowa FC24M for the same, precisely so they can keep up with the high-resolution cameras they are paired with. When you specify a system, the resolution question is not just how many megapixels the camera has. It is whether the lens can actually deliver that resolution to the sensor.

There is a practical version of this worth stating. Work out the smallest feature you genuinely need to detect. Make sure the whole imaging chain, the camera and the lens together, can resolve it. Specifying a resolution the lens cannot support is a very common and very avoidable way to end up disappointed.

Mistake four: forgetting about distortion

The last of the big four is the one that quietly ruins measurement applications. Distortion is the tendency of a lens to bend straight lines. A straight edge near the edge of the image appears slightly curved. Every standard lens has some distortion. For many inspection tasks, a presence check or a simple defect check, a small amount does not matter at all. But if you are measuring, distortion is a direct source of error. A feature measured near the edge of the image will read differently from the same feature measured in the centre.

The mistake is to ignore distortion when the application is a measurement one. People are then puzzled when the readings drift depending on where the part sits in the field of view. There are two ways to handle it properly. The first is to choose a low-distortion lens and, where the software supports it, calibrate out the residual distortion that remains. The second, for the most demanding measurement, is to use a telecentric lens. A telecentric lens is designed so that magnification does not change with distance and distortion is extremely low. That is exactly what precise gauging needs. The high-resolution lens families we supply include low-distortion designs. There are telecentric options for measurement work too. The point is to decide early whether your application is a measuring one. If it is, distortion is something you have to specify for, not discover later.

The four mistakes at a glance

Here are the four together, with what each one actually costs you. The common thread is that all four are decided before you buy. All four are avoidable with the right calculation up front.

Why these mistakes are so easy to avoid

What connects all four of these is that they are specifying mistakes, not engineering failures. None of them happens because the problem was too hard. They happen because a lens was chosen by eye, or by habit, or by whatever was in the drawer, instead of by working out what the application actually needs. That is good news. It means they are entirely preventable. Preventing them costs nothing but a little thought before the order goes in.

The working distance, the sensor coverage, the resolution and the distortion are all things you can determine in advance from the parameters of your application. The part size, the space you have, the smallest feature you need to see, plus whether you are measuring or just checking. Feed those in and the right lens specification falls out. It is far cheaper to get this right on paper than to order the wrong lens, wait for it and find out on the line that the image is not good enough.

Get the numbers right before you buy

The lens deserves the same care as the camera, since it sets the ceiling on everything the camera can do. Guessing the working distance, mismatching the sensor, under-specifying resolution and forgetting distortion are the four mistakes behind most disappointing lens choices. All four come down to doing the calculation before you commit rather than after. A lens chosen properly disappears into a system that simply works. A lens chosen by guesswork is a problem you fight for the life of the installation.

To make the calculation easy, we have built a lens calculator that takes your working distance, your field of view and your sensor, then works out the focal length and the lens specification you need. It is the fastest way to avoid every one of the mistakes above. It is free to use. If your application is tricky, or you would rather have a person check the choice with you, our team is always happy to help. We can prove a lens on your actual parts in our lab before you commit.

Use our lens calculator: get the numbers right

Get in touch: info@clearview-imaging.com | +44 (0)1844 217270

Related: Lens calculator   |   Machine vision lenses   |   Complete Camera Guide   |   Insights Test Lab