Building an optical system with stock optics

Published 15.5.2014

Making the best of building an optical system with stock optical parts.

Several manufacturers can provide pre-fabricated optical parts for very reasonable prices. For a customer they are very attractive, but for the designer they are that sinister oppressive path through a damp dark forest after reading Lovecraft novels - how can anything innovative come out of something as coarse as pre-fabricated parts? Client is demanding the impossible! Or when viewed with a positive attitude, just kicking you forward. Pre-fabricated components design is very attractive and here to stay, so best to make the best of it. Here I design some basic finite conjugate objectives with pre-fabricated parts and see their limitations first-hand.

Since I'm doing just an exploration, I don't have any definite starting specifications. I started with choosing Edmund Optics components that would give me approximately 10x magnification, i.e. a plano-convex spherical lens and an aspherical so that their NA ratio is about 0.1. Furthermore I chose relatively large NA for the image side aspherical, since I suspected I would be improving image quality via throttling the system stop and therefore reducing the image NA. Setting distances manually so that the rays are collimated between the components, I got the following design:

There is diffraction-limited performance at design wavelength (d, 587 nm) for approx. 2° FFOV. However, field and spectral performance is not all that, as was expected for a singlet design. For a localized laser application even this design would perform, though.

Next I tackled the field performance by adding a meniscus lens from Thorlabs. At the expence of the object NA, I increased working distance (and object size). Performance remained pretty much the same, but the design has potential for industrial machine-vision applications now. Magnification also increased.

Next logical step is to tackle the spectral performance by achromatizing components. I changed the first elements to their respective achromats leaving the meniscus as singlet and readjusted the distances, this time automatically with Zemax optimization. Results show remarkedly improved spectral performance with a wider diffraction-limited area. Field of view remained the same while total axial length increased. Image side NA remains the same.

There are certainly some applications where design like this can be useful, but just for comparison's sake I set all dimensions as variables to see how much system would improve with custom components with this starting position. Results show diffraction-limited performance over the whole area inspected (room for widening) over the visual spectral range. The working distance has reduced, but is still over 500 mm, quite generous for industrial applications. The vignetting that plagued the stock design is gone, and the image NA remains 0.2 while object NA of 0.0017 is still usable for machine vision under basic lighting arrangements.

Although it is easier to attain good results with undetermined specifications, end performance suggests stock optics design is not impossible even with more stringent system specifications. Even when custom optics is needed, it is not necessary to assume all components need to be custom. Stock optics design is the best starting point to look which components can be left as pre-fabricated and which are better to manufacture.