Optical design is the process of describing the refracting or reflecting elements in an optical system so that it meets a set of performance specifications. Typically, the performance specifications concern the imaging characteristics of the system, such as the resolution, magnification, numerical aperture, and field of view. Other requirements may include tolerances, size, weight, and cost. The result of an optical design project is typically a prescription or database that lists the materials and shapes of the optical elements required.
A few optical systems that have a limited number of degrees of freedom can be designed using automatic procedures. However, once the complexity of the system or the imaging requirements reach even fairly modest levels, the art of optical design becomes important, and all contests have shown that the best designs are produced by the best designers. Although there has been much ballyhoo about global optimization of optical systems, this does not mean that optical systems can be designed by computers, without human intervention.
Optical design software is numerical optimization software that helps convert a starting design for an optical system into a finished design that meets a given performance specification. The performance specification is described by an error function, often called a merit function. The design software computes changes to the system parameters, or variables, that minimize the error function. Various optimization methods are used to perform the changes: the most widely used is the Damped-Least-Squares method that involves computing the derivatives of the error-function operands with respect to the variables. Other non-derivative, stochastic, and chaotic methods are also used for special problems.
Since part of the optical design process involves computing a function that describes optical performance, optical design software is also used for simulation. The accuracy of the simulation depends on the particular performance attribute being computed. Some characteristics, such as the magnification or field of view, are easily and accurately computed, while others, such as the quality of a complex image, are more difficult to model. It is important to realize that the requirements for simulation error functions are different from those for design error functions. The former must be accurate, while the latter is only required to make a relative quality judgment between one system and another.
Lenses have traditionally been fabricated using methods that involve grinding two surfaces together, which automatically generates a spherical surface. Spherical surfaces, together with the fundamental laws of optics, produce images that contain defects called aberrations. Optical design usually involves balancing contrary aberrations, rather than removing them. New materials and aspheric fabrication techniques have expanded the possibilities for optical design, but the general process remains one of balancing rather than removing aberrations.
The required nonlinear optimization and the difficulty of specifying performance requirements in a simple yet complete manner make optical design both an art and a science. A skilled optical designer must use a combination of scientific knowledge and judgment to produce an optimized design. Modern computer technology has greatly simplified the process of managing the data and balancing the aberrations, providing optical design capability to a much wider community than before. In addition, graphical interfaces make programs like OSLO much easier to use than traditional console applications.
Traditional optical design is based on ray tracing or aberration theory. Ray tracing is essentially a sampling technique in which data for a few rays are extrapolated to indicate the performance of an entire system. Aberration theory provides a different type of sampling, in which low-order performance coefficients are balanced with high-order performance coefficients to establish overall performance.
The biggest problem in optical design is the construction of a suitable error function that requires only a few data points, but nevertheless provides an accurate indication of actual performance. It is not generally possible to set up a single error function that will do this for an arbitrary system, without unacceptable loss of computational efficiency. The error function must be customized for the problem at hand. Depending on the facilities that are provided for setting up customized error functions, in combination with the talent and experience of the user, this may be a greater or smaller limitation of a particular optical design program.
A more fundamental limitation of optical design software concerns the representation of continuous physical data in a computer. Even though optical design software uses arithmetic with up to 15 digits of precision, the need to balance contrary aberrations determined by nonlinear equations complicates the optimization of many systems because of numerical effects.
The Learning section of this web site has an extensive article on optical design software, and also a list of general reference books in optical design. If you would like to learn about optical design software, and would like to see what OSLO can do, please download OSLO EDU and try it on some simple problems. OSLO EDU is a complete optical design program for systems containing up to 10 surfaces, and is intended for educational use (either formal or self-improvement).