More than a year ago, I wrote a post on lens adapter tolerance, bemoaning the fact that all the adapters that I’d tested were too short. I followed that up with a post about how, at least in the case of Novoflex, this was on purpose. As I continued to purchase adapters from various manufacturers, all were about 0.5mm too short, which made the readings on the focus scale invalid, threw off parfocality in zoom lenses, and rendered less effective lenses with internal focusing and/or floating elements.
The week before last, I received an adapter from Kipon which allowed me to use my 5 cm f/2 Nikkor-H lens, which was originally designed for Nikon S-mount cameras, on my Sony alpha 7 cameras. For that lens and some others, the helicoid was in the camera body, and thus any adapter has to have such a helicoid so that the lens can be focused. I noticed that the distance scale on the Kipon adapter was accurate. That meant that the adapter wasn’t too short.
I wondered if the tolerances were so tight on this adapter because it had the focusing mechanism built in. In order to gain some insight into the matter, I ordered a Kipon adapter that was set up for mounting a Leica M-mount lens on a Sony NEX, E, or FE mount body. It came yesterday. Today I put it on a Sony a7II, and attached a Leica 50mm Summicron-M ASPH. The focusing scale was accurate. Infinity focus occurred just barely before the infinity stop on the lens, just as it does when that lens is mounted to a real Leica M9 or M240.
So, based on a sample space of two, it’s possible that there’s at least one manufacturer that doesn’t make their adapters deliberately way too short. Hallelujah! I’ve ordered Kipon Nikon F to NEX and Leica R to NEX adapters and I’ll let you know how tight the tolerances are when they arrive.
The whole thing has gotten me thinking in more quantitative terms about adapter tolerance. I’ll share some of those thoughts with you.
First, let’s consider the tolerance situation when a lens is mounted on the camera for which it was designed, with no adapter in the picture. (Note: for this and the ensuing tolerance discussion, I’m going to assume that tolerances are absolute, that if I say the tolerance is 10mm +/- 0.5mm, that means that the shortest the part will be is 9.5mm and the longest it will be is 10.5mm.)
Let’s say that the camera and lens manufacturer sets the flange focal distance to some nominal quantity that we’ll call ffd. Now the manufacturer assigns a tolerance to the cameras ffd, say +/- c and to the lens, say +/- l. Then the camera and lens manufacturer has to design the lens so that it focuses far enough beyond the infinity mark on the lens so that the image will be in focus if the actual worst case flange focal distance of lens and camera is ffd + c + l. You will see the greatest effect on the focusing scales of wide angle lenses.
It has been my experience for manual focusing lenses with precisely marked distance scales, with the exception of one Zeiss 35mm f/2 Biogon that I own but hardly ever use, that the distance scales of these lenses is accurate enough to achieve proper focus wide open by measuring subject distance with a tape measure and setting that value with the focusing ring.
In messing with a bunch of lenses and their native bodies, I estimate that c and l add up to less than 0.05 mm, maybe a lot less. I used Leica M cameras and lenses, Nikon F-mount cameras and Zeiss F-mount lenses for my testing. 0.05 mm may sound small, but converted to inches it’s 0.002. Around the metal shops at Hewlett-Packard (the last ones with which I have any personal experience), +/- 0.001 inches (usually called a mil) for parts the size of a lens adapter, wasn’t considered truly precision machining, but the sort of thing you’d give to the new guy. I should note that the sensors on cameras can be, and often are, shimmed into tolerance, so that the machining requirements are lessened.
Here’s the worst-case infinity focusing error for the camera and the lens at the end of their tolerance band in the direction that would put the lens closest to the focal plane:
Now, let’s consider adding an adapter into the mix. If we design the adapter’s nominal thickness as the difference in flange focal distances between the camera for which the lens was designed and the camera to which the adapter will let it attach, on average the distance scales on our lens will be just as accurate as they were before the adapter appeared on the scene, but there is the possibility that the adapter will keep a particular camera/lens combination which was out of tolerance in the worst direction from focusing to infinity.
We can keep that from happening by specifying the tolerance for the adapter thickness as +/- a, and making the nominal adapter thickness shorter than the difference in flange focal distances by a. Now, the lens will always focus to infinity if the manufacturers of the camera and lens had similar ideas about what the tolerances should be.
So, what should a be? Let’s assume the adapter thickness is achieved through precision surface grinding. A little Internet research shows that standard surface grinding tolerances are about +/- 0.0003 inches.
Then our nominal adaptor would be too short by 0.0003 inches, and, paired with a nominal camera/lens combination, would produce focusing errors like this:
If the worst-case short adapter were teamed with the worst-case short camera and lens combination, the focusing error would look like this:
Let’s see. An 18mm lens could be off so far that when it’s focused to infinity, the marking on the barrel would say it was focused to about 150 feet, That’s certainly within the tolerance allowed by the depth of field of an 18mm lens, even if it’s fast.
So why do we see adapters from everybody but Kipon — at least everybody I’ve tried: Novoflex, Metabones, Fotodiox, MTF, adn some others — short by so much?
It’s a puzzlement to me. Is it hard for them to hold the kind of tolerances I’m positing above? Are they biasing their design specs so that truly out-of-tolerance lenses, ones that wouldn’t focus to infinity on their native bodies, will focus to infinity on adapted bodies?
I’d appreciate input from anyone who has a good understanding of modern machining tolerances.