Optical system design probably means different things depending on who you ask. If you ask me, it is more of a process than a task. A process is something you would continuously execute, feeding insights back to the different stages of development. A task would be something like developing a specific system or module to meet specific requirements.

Limiting the design process to a set of tasks limits the benefits of feedback and continuous learning, one of which is how well the results match the performance predictions from the design process.

Closing this loop can only be done by measurements of specific optical performance parameters. It is not unusual that this skipped over. The optical system is just a cog in a greater machine and if the machine performce once, parts are assumed to be performing as well. In my experience, that is not always, perhaps not even usually, the case. Complex systems can hide underperforming parts and just because one particular device or machine works does not automatically mean that all parts perform as designed. For complex systems, the law of large numbers comes into play. Unless the parts are over-specified, when we build enough systems we will find that too many fail to reach the goal, and in the process, a whole lot of resources were spent chasing too difficult targets because one sub-system was not up to par.

In the high-performance business, and by that I usually think of optical systems that perform well beyond the requirements of the classical diffraction limit, optical system design must be a chain, perhaps I should even call it a necklace because of the feedback to design methodology and tools, where each link provides a traceable step towards the desired outcome.

But there is a practical distinction to what is optical system design and what it is too often confused with, lens design. Lens design starts with a set of requirements such as size, type of imaging, wavelength range, field angles or image size, and optical quality over the range of these variables. System design is to determine what all these parameters should be. Once the ray-tracing specialist (Zemax/OpticStudio, Code V and so on) knows that, he can do a fantastic job. Without it, it’s all guessing or best effort. How do we know how much spherical aberration, field curvature or distortion is allowed in order to meet the specified customer requirements? Are there any requirements on coherence, spatial or longitudinal. But let’s make one last deep-dive into the mathematics. How should we describe the optical performance of the imaging system itself. Is it appropriate to use the Zernike polynomials? Many would say, if the pupil is circular then the Zernike set is the right one. Sure, if it is a telescope, projector or camera lens you are making? Absolutely. However, if your performance requirements cannot be directly traced to wavefront variance, then the Zernike set, as much as we may like, or even love it, is probably not the correct one. All of that is the optical system designers job to find out.