Telecentric Lenses: "Is it all Hype?"

We get a lot of questions about telecentric lenses. It's really no wonder; there are some counter-intuitive aspects to them. One thing that is certain; they have a place in machine vision. This article will address the telecentric lens and especially how to recognize the suitability of a telecentric lens to your application.

We hope to dispel the myths and expound on the merits of these fascinating optical devices. We'll begin with a description of a telecentric lens, from a technical point of view.

Most of us are already familiar with a variety of optical systems. A telescope allows us to see things that appear very small because they are so very far away. Telescopes are often described by the diameter of their primary optical element. A microscope shows us things that are both very small and very close. Microscopes are often described in terms of their magnifying power. The general family of photographic optics creates an image of our friends and family on a small piece of film, or a computer chip. The lenses most often used in machine vision are members of this family. In all of these systems the chief rays of light from the object form a cone of light that is collected by the lens system. The light is then manipulated so that it will create the image we want. The manipulation of the light bundle by the lens system is often summarized in terms of the focal length, f/# and other specifications of the lens.

The Science behind Telecentric Lenses

Let us consider an ordinary lens setup for imaging from a paraxial point of view. In this case we have shown three object planes labeled O1, O2, O3 respectively.

Figure 1: Typical Layout of Imaging System

There are a number of things that can be seen by this simple sketch. The ray bundle consists of the chief ray (the ray in the center of the three) and this Chief Ray is at an angle to the optical axis. The ray that is parallel to the axis is called the marginal ray.

Examining the Chief Ray (highlighted in blue) we notice that it is a different height relative to its distance from the lens. So for equal sized objects, objects closer to the lens will appear bigger (the Chief Ray is closer to the optical axis) and objects further away appear smaller (the Chief Ray is further from the optical axis).

Now consider what occurs if we were to simply place a stop behind the lens at the rear principal focus of the lens, then :

Figure 2: Basic Principles behind telecentric lenses

Now in this case, notice that the Chief Ray has become parallel with the object axis. A critical thing to notice is that the height of the ray relative to the axis of the system remains constant, regardless of the distance from the lens system itself. So objects remain the same size (i.e. constant magnification) throughout the range. As we will discuss below, they will lose focus just like an ordinary lens as the object moves in and out of the focal plane, but the centroid of the imagery remains the same throughout the different object distances. This is extremely useful in gauging or size dependent machine recognition tasks.

This basic property can be extended for more complex lenses. If a multi-element system is used as in real world lenses, care is taken to place the exit pupil of the sensor lens at the rear focal plane of the front lens as shown above. The next group of elements is situated such that the entrance pupil is co-located with the exit pupil of the preceding group.

Another interesting consequence pointed out by Watanabe and Nayar (1996) that the effective f/# does not change with object distance since the aperture is not varying. Thus for various object distance (within the depth of focus) a falloff in light intensity is not observed.

Another less desirable consequence is also readily seen by Figure 2. In order to image a given object, then the first element of the system must be slightly larger than the object itself to allow for the light collection and telecentric condition. In other words there is a realistic size limitation of the ability to image. As a result often the lenses get to be quite large and heavy and hence have to be adequately supported.

The Truth about Telecentric Lenses

Fact One:

Telecentric lenses have no parallax distortion. This might just be the point of greatest interest to machine vision professionals. This lets you see the bottom of holes regardless of their position in the field of view.

Fact Two:

Telecentric lenses are ideal for measurement applications. The telecentricity of this class of lenses makes them ideal for a wide range of measurement applications. An object will appear to be a consistent size regardless of its location in the field of view in a properly designed telecentric system. This is especially useful when linear arrays are used to "stitch" together multiple images into a cohesive scene. In this way a line scan camera can create an accurate image of infinite length without distortion. The use of telecentric lenses with linescan cameras is by far the largest application area that we have seen.

Fact Three:

Telecentric lenses allow variations in working distance. Within the depth of field, a telecentric lens will show no magnification error. This makes setup and calibration less complex.

Fact Four:

There are different types of telecentric lenses. Lenses can be telecentric in object space such as used in gauging applications. These are the most common. They can also be telecentric in image space such as used for depth from defocus applications. In some cases they can be telecentric in both spaces simultaneously such as in a telescope. Furthermore, there are a few telecentric zoom lenses available as well. However, care should be taken when setting up commercial optics as a telecentric system as the pupil of a zoom lens will tend to move with the zoom action and hence the telecentric condition can be violated.

Telecentric Myths

Don't be fooled by claims about telecentric lenses. Here are some common myths.

Myth One:

A telecentric lens offers greater depth of field. Nope. Telecentric lenses are limited to working within the laws of physics, just like traditional photographic optics. Depth of field is increased by decreasing aperture, and many telecentric lenses have smaller apertures than you would expect. That may explain the persistence of this myth.

Myth Two:

A telecentric lens is the only way to make accurate measurements. Not at all. Given a known parallax, the image can be mapped and corrected. Also a given position or marker can be setup in a MV system such that the image always is taken from the same point. Thus even if there is a distortion in the lens, since identical images are "learned" from the same point in object space, the effective distortion is accounted for.

Myth Three:

A telecentric lens is machine vision's "magic bullet." Sorry. It just isn't that simple. A telecentric lens will not correct illumination problems, or increase your manufacturing yield. The entire vision system must be designed with the same care as COI's custom lenses.

Myth Four:

Telecentricity is the only important lens characteristic. Wrong. The actual lens chosen still depends upon the resolution, distortion, and other lens specifications that affect the application.

Conclusions

The use of telecentric lenses in machine vision has grown quickly in the last few years. And with good reason. Many applications can effectively take advantage of telecentricity. We hope that this article will help to clarify when a telecentric lens is indicated.

You will find LensZilla, the largest telecentric lens in the World, here. The data sheet for another COI-designed and manufactured Telecentric lens is also available. You will find more information about our most popular telecentric lens in this data sheet. Computar Tec-55 information

Contact COI for your telecentric lens needs. We carry stock lenses from the key manufacturers, and will design and manufacture a telecentric lens for your application.

About the Author

Dr. Jonathan Kane joined Computer Optics Inc in 1996 as Director of R&D, and is now the President. Prior to that he was a staff scientist with the Air Force Research Laboratory. He has published a variety of articles and holds a series of patents in: Machine Vision, Image Processing, Ferroelectric/VLSI display devices, Optical Processing, Photorefractive Crystals, Phase Conjugation, Atomic Phenomena, Fourier Processing, Neural Networks, and Optical Logic. His current interests include Prototype electro-optical construction as well as Optical Lens Design and Optical Coating design. Questions or comments about this article can be sent to him at COI.