Subjectivity in engineering design

As is often the case, the impetus for me to take keyboard in hand to click out this post is an exchange with a fellow photographer over the ‘net. In this case, the conversation went something like this:

Someone else: “I am not a fan of MTF graphs.”

Poster: “I look at real-world shots from objective reviewers mainly. Also, before purchasing, I would take a few test-shots myself if possible and process and view at home. There is so much that is not shown in an MTF graph.”

Me: “Sounds like the answer is ‘none of them.'”

Poster: “There are no objective ‘metrics’ for photographic quality as far as I know.”

Me: “If that’s true, what is a lens designer trying to optimize?”

Poster: “The subjective aspects of the output that he likes (or is told to optimize by his company). “

I am not a lens designer. But I don’t think it works that way, based on my long career as a design engineer. Engineers don’t like subjective boundary conditions. It makes it hard to optimize. If there are subjective elements to the design criteria, engineers like to look for ways to make them objective.

Here’s an example. Before electric typewriters, the feel of a typewriter keyboard was determined by the mechanical design of the moving parts of the machine. When electric typewriters came along, the engineers found that, within reason, they could make the feel of the keyboard whatever they wanted it to be. The obvious question was, “What should it feel like?” The folks at IBM made a bunch of keyboards with differing characteristics, rounded up a group of professional typists, and conducted tests to find out what they liked best in terms of keyboard feel. They varied the stroke length, the spring tension, the actuation point, the force vs key depression curves, and finally came up with a set of specifications for they keyboard that most users like the best. Those specs were objective: quantitative, measurable, repeatable. The engineers’ task was to devise keyboards with the feel defined by the specs that were inexpensive, durable, and manufacturable. The engineers did their jobs. The resulting keyboards became the standard of the industry. Professional typists loved them.

Over time, the users of keyboards broadened to include more and more people who were not professional typists. Now just about everyone uses keyboards, and they have different requirements. In addition, keyboards are often used in cost-conscious and space applications where the original IBM feel is just too expensive and too big. Enter cheap spongy switches with short travel. That was a decent solution for the masses — although I personally hate those kinds of keyboards. But many users wanted longer travel keyboards with various kinds of feel, and the industry responded with a wide variety of mechanical and magnetic switches, all with different feel. Each of those switches has been specified in objective terms. I happen to be a fan of the original IBM keyboard feel, but now my favorite keyboard is a Hall Effect device made by Keychron. It has adjustable actuation points and even is configurable to activate different functions at different depression depths.

To the uninformed user, the only way to find out if you like a keyboard or not is to try it. That’s not a very efficient way to make a purchasing decision. Users who are a bit more savvy learn the characteristics of keyboards that they like, and purchase boards that are objectively similar to the ones they like. The engineers designing all those keyboards presumably design the feel to the numbers. I doubt if a product manager tells an engineer to just go off and design a keyboard that feels good to him.

I expect that it’s the same with lens design. The lens designer has a double armload of objective parameters: cost, size, weight, flare, spherical aberration, astigmatism, coma, LoCA, LaCA, spectral response, light falloff, distortion, far-OOF point spread functions, radial and tangential MTF, optical passband, transitions from in to out of focus, and many more. All those parameters are measurable and can be simulated, so that the lens designer can see the result of her design decisions before actually constructing the lens and hanging it on a camera.

But it’s not just cut and dried. In general all the parameters are affected by the lens design, and it’s not possible to obtain a perfect design. Enter the world of tradeoffs. The designer, possibly in conjunction with the product manager, must assign weights to all of the parameters to obtain a scalar value to be maximized or minimized. The picking of those weights is subjective.

But that doesn’t mean that prospective users of a lens can’t gain a lot of valuable information about whether that lens will suit their needs from looking at objective metrics. It’s more complicated than picking a keyboard, there’s a lot more data to be sifted through, and much of that data is not published by many of the lens vendors, but, just like a keyboard, there are objective metrics that can in the hands of a sophisticated consumer, contribute mightily to a well-informed decision.

Here is an opportunity for me to toot my own horn a bit. When I test lenses, I provide a wealth of detail about many metrics that I consider important. I think that examining my lens tests can go a long way to informing people about whether or not a particular lens is for them. I have been backing down somewhat on the quantitative testing, because it’s difficult and I think that much of it serves to confuse some people, but I do try to provide information about the lens parameters that I consider important.