Using stock optical components for a project is an attractive prospect, but knowing which components to use can help navigate the most common pitfalls.
Stock optical components are very attractive option for a developer. You get parts ASAP, keep prototyping costs down and one gets fitting optomechanical parts to go, and you can do all by yourself as well. Then again, there are whole factories spewing out same parts to your competitor, with same instructions and howto's. So it comes down to this: what SPECIAL can you do with stock optical parts? Turns out, quite a lot. However, they might need an optical engineer to pull them off. I'll go through some examples of common tripping points of stock optical components, and a short description on how an optical designer can overcome them.
Stock optics manufacturers (and basic books on optics, even Wikipedia) give enough information on how to construct your own optical system to diffraction-level performance. However this applies to on-axis image only, meaning if you want wider field than a degree or few, you might be disappointed. A proper meniscus lens inserted in the right place can widen your field, but a professional design software can shave off weeks of trial and error in the workshop, not to mention component costs it causes.
Bandwidth is also a good question. All stock optics have a design wavelength (just the one) to specify focal and back focal length. This doesn't mean others or even wider bandwidths are impossible to utilize - it simply tells in what wavelength the focal length and back focal length measurements were made. Unfortunately it will not tell anything about the other wavelengths or how it behaves with a wider bandwidth. The index of refraction of a glass falls faster at UV than at IR, so be very precise at UV, use achromatic lenses at visual and relax at IR.
Tolerance analysis is crucial when assembling high resolution systems. All you have as an indication about the resolution power is the final image plane (and intermediate image planes, if any). That tiny spot of a projected test point source (you remembered to make one, right?) can reveal the state of the system via i.e. Star Testing, but it is a qualitative test that requires very high level of experience from operator. Simpler tests are available, but require further apparati. All the testing in the world would not save you from trying to desperately find out the offending variable, a huge handicap in scheduled projects. The simplest solution is a tolerance analysis that shows you the trouble spots in advance. It considers changes in all variables and assigns lower and upper tolerance limits to all considered variables that still achieves performance you specified. Tolerance limit calculations are just raw Monte Carlo calculations, an optical designer's software does them in time but without further effort from you. Tolerance analysis is always included in the optical design. Same points are valid for coatings and transmission analysis as well.
Finding the correct focal lengths is really not difficult at all. However, the calculations are most likely done paraxially, and that means the thickness of the lens, shape factor or the characteristics of the glass are not considered at all. This causes third and higher order aberrations that will sink the resolution power quite fast. Also messing up good ideas, stock optical components come in many forms: you can get an f=100mm lens in Plano-convex, Double Convex, Aspheric, Best Form, Meniscus and Hybrid forms, in the approximate order of complexity. How much time would it take to go through all the theory behind them all to determine which one is right for you? In a nutshell: Plano-convex is usually a good idea for infinite conjugate locales and double convex for ones with finite conjugate. Aspherics are good for removing spherical aberrations but perform poorly with field height, so the Best Form lens might be better for those. Menisci can adjust wavefronts and fields, and expensive hybrids are used for serious relative aperture and illumination. Adding to the complexity, two (or three?) simpler form lenses could do the trick of one special form with fifth of the price. It all comes down to minimizing the angles of incidence and exitance. These are hard calculations, and if wanted in timely fashion, best use an expensive professional software - the kind optical designers have already acquired and mastered.
The only optical system that wont benefit from optimization is an optimized system. Optimization such as DLS (or Levenberg-Marquardt, if you want college-period nightmares revived) is the work horse of optical design. In the simplest form, you specify what you want in the demerit function, and the DLS will optimize and tell you the new demerit function value, i.e. if your requirements are possible with the chosen optical layout. Simple as A-B-C, if you have the software and skills for it (guess who already has?). Operating the DLS is easy, configuring the demerit function is not. Demerit functions are very susceptible to "Doing what was told, not what was wanted" problem. It has a philosophy that incidentally needs deeper knowledge of the DLS algorithms and relations between aberrations and variables. This can be easily considered the steepest learning curve of optical design. Of course there are tricks and shortcuts to it.
Using stock optics is by no means discouraged. The purpose of this article is to point out possible tripping points when pushing the limits. Asking for a free quotation from an optical designer could very well reveal pleasant time-saving information.