Is A DSLR APS-C Camera More Likely To Create Creamier Bokeh than A Mirrorless APS-C?

I kind of hate to answer because I am leaving for Thanksgiving vacation and won't be around for follow up posts. But the question is really interesting so I will write a longer than average post.

In theory, there is no reason you cannot get just as good of bokeh with short or a long lens registration distance. But in practice, things are more difficult.

In lens design, there is a thing called the symmetric principle. When you can make a lens symmetric from front to back, you tend to get better performance. If you look at cutaway drawings of lenses, you often will notice an overall symmetric appearance. The symmetric principle also applies to the ratio between the distances of lens to object and lens to image. A ratio of one is ideal; everything gets really easy then. But when that ratio gets really high, it becomes more difficult to achieve design goals. Camera lenses almost always work with really high ratios. Short registration distances push that ratio to higher values.

With tougher design goals, more glass elements, tighter manufacturing tolerances, aspheric surfaces or expensive glasses are needed.

When a lens is designed, the final tweaks are done by a computer according to a merit function. A merit function specifies lens performance on axis, various distances off axis, at different wavelengths, allowed amounts of distortion, focus distances or just about anything you can imagine. One of the important things to consider is what is the performance is when the image is slightly, or a lot out of focus. This gets put into the merit function too. When you hear that a lens has great bokeh, you can be sure that it was not that way by accident. It was designed for that characteristic.

The particulars of merit functions that different manufacturers use are considered trade secrets. Very little of use leaks out about them. They reflect a lot of the market research and subjective evaluations that companies have built up over the years to define what a good lens is. But some generalities are known. For instance, Leica lenses show a slow decline in MTF as spatial frequencies increase. But some other manufacturers have lenses that hold up a high MTF value to some frequency but then go to pieces rapidly after that. It all goes back to what the merit functions were emphasizing.

I remember reading a post a while back where someone said they had spent a lot of effort looking at published MTF curves to glean an idea of what lenses would have good bokeh and had found no correlation at all. It is not surprising. The MTF curve is published for best focus. It tells you nothing about the MTF out of focus. That is why lens design programs have a "Through Focus MTF" function. If you write your merit function to emphasize great peformance at exactly the best focus, you typically sacrifice performance for the out of focus conditions. (This is one of the more difficult things for older lens designers to get across to the designers right out of school. It typically takes about five years of experience interacting with final users before they really get it and come back and say,"You know, out of focus performance is really important. The doctor has to be able to recognize the XXX while he getting the YYY into position so he can do the ZZZ." Of course you feel like saying "Yes, we have been telling you that," but we actually say "That is a really good observation.")

It is a common trick in optical design, like in aerospace or medical systems where I typically work, that when we have difficulty meeting the performance goals, we go back to system requirements and see if we can get an allowed increase in distance between the lens and the glass. Usually you can. Often it just takes finding a camera that is shorter so everything can fit into some box. System requirements are usually dreamed up by people who have no idea what the system design tradeoffs are, but for some reason, they like to specify the camera. So it is our job to communicate and work things out.

Camera lenses are a whole different world than where most of my real professional experience rests. But lens designers sometimes refer back to a comment Warren J. Smith made in one of his optical design textbooks: "It is remarkable how few useful general principles there are to master in lens design." (not an exact quote) So something like the symmetric principle is something that actually gets referred to pretty frequently.

The most important thing of course is what do you see with real lenses and real cameras. A few days ago I needed some reference data, so I took my A65 with DT 2.8/30 Macro lens and my NEX7 with E 3.5/30 Macro lens and compared depth of focus and bokeh at F/4 and object distance 24 inches. The performance was very similar. I was a little bit surprised, but not terribly. Prime lenses in general come pretty close to the theoretical best performance, and that would include bokeh considerations. The E mount lens was significantly more expensive though.

I also did some comparisons with the 18-55 mm kit zoom lenses for both cameras. Unfortunately, I seem to have gotten one of the bad E mount kit lenses, so it is not worth making comparsions. With the 18-55 E mount lens photographing a field of USAF resolution targets, there is a patch of fuzzy covering the center left quarter of the camera, but recovering at the edge of the field. It doesn't really matter in the grand scheme of things. I use my NEX7 with a Tamron 18-200 lens almost all the time and that lens is fine. I guess I could return the bad lens, but is useful to me professonally to keep a few clinkers around when discussions go that way.

I also saw that my A65 has a fair amount of forward focus with the macro and kit zoom on autofocus. Almost, but not quite enough, for me to consider picking up an A77 now that those are on sale. (The A77 is a bit big for my taste.) But if you are thinking of issues like bokeh, I would not overlook the fact that it is better to have your object in the center of the depth of field and A77 can manage that a lot better than the A65 on autofocus. The A77 has microfocus adjust but the A65 does not. Otherwise the NEX cameras with contrast detection are going to be doing a better job on getting the object in the center of the depth of field so you might even say a NEX camera is better for bokeh than an A65, but not an A77.

(of course for artistic reasons, you might adjust things a bit differently, but in general I think most of the time the photographer intends to put the object in the center of the depth of field.)